diff --git "a/axquant_cuda_plan.json" "b/axquant_cuda_plan.json" new file mode 100644--- /dev/null +++ "b/axquant_cuda_plan.json" @@ -0,0 +1,9613 @@ +{ + "activation_dtype": "bfloat16", + "activation_precision": 16, + "algorithm": "rtn", + "allocations": [ + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "lm_head", + "scale_group": null, + "shape": [ + 248320, + 4096 + ], + "source_file": "model-00001-of-00004.safetensors", + "tensor_name": "lm_head.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "embedding", + "scale_group": null, + "shape": [ + 248320, + 4096 + ], + "source_file": "model-00001-of-00004.safetensors", + "tensor_name": "model.language_model.embed_tokens.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00001-of-00004.safetensors", + "tensor_name": "model.language_model.layers.0.input_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 32 + ], + "source_file": "model-00001-of-00004.safetensors", + "tensor_name": "model.language_model.layers.0.linear_attn.A_log" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 8192, + 1, + 4 + ], + "source_file": "model-00001-of-00004.safetensors", + "tensor_name": "model.language_model.layers.0.linear_attn.conv1d.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 32 + ], + "source_file": "model-00001-of-00004.safetensors", + "tensor_name": "model.language_model.layers.0.linear_attn.dt_bias" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 32, + 4096 + ], + "source_file": "model-00001-of-00004.safetensors", + "tensor_name": "model.language_model.layers.0.linear_attn.in_proj_a.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 32, + 4096 + ], + "source_file": "model-00001-of-00004.safetensors", + "tensor_name": "model.language_model.layers.0.linear_attn.in_proj_b.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 8192, + 4096 + ], + "source_file": "model-00001-of-00004.safetensors", + "tensor_name": "model.language_model.layers.0.linear_attn.in_proj_qkv.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 4096, + 4096 + ], + "source_file": "model-00001-of-00004.safetensors", + "tensor_name": "model.language_model.layers.0.linear_attn.in_proj_z.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 128 + ], + "source_file": "model-00001-of-00004.safetensors", + "tensor_name": "model.language_model.layers.0.linear_attn.norm.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 4096, + 4096 + ], + "source_file": "model-00001-of-00004.safetensors", + "tensor_name": "model.language_model.layers.0.linear_attn.out_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": null, + "shape": [ + 4096, + 12288 + ], + "source_file": "model-00001-of-00004.safetensors", + "tensor_name": "model.language_model.layers.0.mlp.down_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.0.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00001-of-00004.safetensors", + "tensor_name": "model.language_model.layers.0.mlp.gate_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.0.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00001-of-00004.safetensors", + "tensor_name": "model.language_model.layers.0.mlp.up_proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00001-of-00004.safetensors", + "tensor_name": "model.language_model.layers.0.post_attention_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00001-of-00004.safetensors", + "tensor_name": "model.language_model.layers.1.input_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 32 + ], + "source_file": "model-00001-of-00004.safetensors", + "tensor_name": "model.language_model.layers.1.linear_attn.A_log" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 8192, + 1, + 4 + ], + "source_file": "model-00001-of-00004.safetensors", + "tensor_name": "model.language_model.layers.1.linear_attn.conv1d.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 32 + ], + "source_file": "model-00001-of-00004.safetensors", + "tensor_name": "model.language_model.layers.1.linear_attn.dt_bias" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 32, + 4096 + ], + "source_file": "model-00001-of-00004.safetensors", + "tensor_name": "model.language_model.layers.1.linear_attn.in_proj_a.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 32, + 4096 + ], + "source_file": "model-00001-of-00004.safetensors", + "tensor_name": "model.language_model.layers.1.linear_attn.in_proj_b.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 8192, + 4096 + ], + "source_file": "model-00001-of-00004.safetensors", + "tensor_name": "model.language_model.layers.1.linear_attn.in_proj_qkv.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 4096, + 4096 + ], + "source_file": "model-00001-of-00004.safetensors", + "tensor_name": "model.language_model.layers.1.linear_attn.in_proj_z.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 128 + ], + "source_file": "model-00001-of-00004.safetensors", + "tensor_name": "model.language_model.layers.1.linear_attn.norm.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 4096, + 4096 + ], + "source_file": "model-00001-of-00004.safetensors", + "tensor_name": "model.language_model.layers.1.linear_attn.out_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": null, + "shape": [ + 4096, + 12288 + ], + "source_file": "model-00001-of-00004.safetensors", + "tensor_name": "model.language_model.layers.1.mlp.down_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.1.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00001-of-00004.safetensors", + "tensor_name": "model.language_model.layers.1.mlp.gate_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.1.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00001-of-00004.safetensors", + "tensor_name": "model.language_model.layers.1.mlp.up_proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00001-of-00004.safetensors", + "tensor_name": "model.language_model.layers.1.post_attention_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.10.input_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 32 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.10.linear_attn.A_log" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 8192, + 1, + 4 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.10.linear_attn.conv1d.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 32 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.10.linear_attn.dt_bias" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 32, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.10.linear_attn.in_proj_a.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 32, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.10.linear_attn.in_proj_b.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 8192, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.10.linear_attn.in_proj_qkv.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 4096, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.10.linear_attn.in_proj_z.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 128 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.10.linear_attn.norm.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 4096, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.10.linear_attn.out_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": null, + "shape": [ + 4096, + 12288 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.10.mlp.down_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.10.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.10.mlp.gate_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.10.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.10.mlp.up_proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.10.post_attention_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.11.input_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": null, + "shape": [ + 4096, + 12288 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.11.mlp.down_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.11.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.11.mlp.gate_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.11.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.11.mlp.up_proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.11.post_attention_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 256 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.11.self_attn.k_norm.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": "model.language_model.layers.11.self_attn.qkv_proj", + "shape": [ + 1024, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.11.self_attn.k_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 4096, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.11.self_attn.o_proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 256 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.11.self_attn.q_norm.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": "model.language_model.layers.11.self_attn.qkv_proj", + "shape": [ + 8192, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.11.self_attn.q_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": "model.language_model.layers.11.self_attn.qkv_proj", + "shape": [ + 1024, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.11.self_attn.v_proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.12.input_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 32 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.12.linear_attn.A_log" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 8192, + 1, + 4 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.12.linear_attn.conv1d.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 32 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.12.linear_attn.dt_bias" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 32, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.12.linear_attn.in_proj_a.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 32, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.12.linear_attn.in_proj_b.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 8192, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.12.linear_attn.in_proj_qkv.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 4096, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.12.linear_attn.in_proj_z.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 128 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.12.linear_attn.norm.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 4096, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.12.linear_attn.out_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": null, + "shape": [ + 4096, + 12288 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.12.mlp.down_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.12.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.12.mlp.gate_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.12.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.12.mlp.up_proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.12.post_attention_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.13.input_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 32 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.13.linear_attn.A_log" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 8192, + 1, + 4 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.13.linear_attn.conv1d.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 32 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.13.linear_attn.dt_bias" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 32, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.13.linear_attn.in_proj_a.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 32, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.13.linear_attn.in_proj_b.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 8192, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.13.linear_attn.in_proj_qkv.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 4096, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.13.linear_attn.in_proj_z.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 128 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.13.linear_attn.norm.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 4096, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.13.linear_attn.out_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": null, + "shape": [ + 4096, + 12288 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.13.mlp.down_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.13.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.13.mlp.gate_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.13.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.13.mlp.up_proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.13.post_attention_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.14.input_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 32 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.14.linear_attn.A_log" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 8192, + 1, + 4 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.14.linear_attn.conv1d.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 32 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.14.linear_attn.dt_bias" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 32, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.14.linear_attn.in_proj_a.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 32, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.14.linear_attn.in_proj_b.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 8192, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.14.linear_attn.in_proj_qkv.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 4096, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.14.linear_attn.in_proj_z.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 128 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.14.linear_attn.norm.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 4096, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.14.linear_attn.out_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": null, + "shape": [ + 4096, + 12288 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.14.mlp.down_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.14.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.14.mlp.gate_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.14.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.14.mlp.up_proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.14.post_attention_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.15.input_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": null, + "shape": [ + 4096, + 12288 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.15.mlp.down_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.15.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.15.mlp.gate_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.15.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.15.mlp.up_proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.15.post_attention_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 256 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.15.self_attn.k_norm.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": "model.language_model.layers.15.self_attn.qkv_proj", + "shape": [ + 1024, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.15.self_attn.k_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 4096, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.15.self_attn.o_proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 256 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.15.self_attn.q_norm.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": "model.language_model.layers.15.self_attn.qkv_proj", + "shape": [ + 8192, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.15.self_attn.q_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": "model.language_model.layers.15.self_attn.qkv_proj", + "shape": [ + 1024, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.15.self_attn.v_proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.16.input_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 32 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.16.linear_attn.A_log" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 8192, + 1, + 4 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.16.linear_attn.conv1d.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 32 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.16.linear_attn.dt_bias" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 32, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.16.linear_attn.in_proj_a.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 32, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.16.linear_attn.in_proj_b.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 8192, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.16.linear_attn.in_proj_qkv.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 4096, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.16.linear_attn.in_proj_z.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 128 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.16.linear_attn.norm.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 4096, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.16.linear_attn.out_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": null, + "shape": [ + 4096, + 12288 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.16.mlp.down_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.16.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.16.mlp.gate_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.16.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.16.mlp.up_proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.16.post_attention_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.17.input_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 32 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.17.linear_attn.A_log" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 8192, + 1, + 4 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.17.linear_attn.conv1d.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 32 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.17.linear_attn.dt_bias" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 32, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.17.linear_attn.in_proj_a.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 32, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.17.linear_attn.in_proj_b.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 8192, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.17.linear_attn.in_proj_qkv.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 4096, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.17.linear_attn.in_proj_z.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 128 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.17.linear_attn.norm.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 4096, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.17.linear_attn.out_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": null, + "shape": [ + 4096, + 12288 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.17.mlp.down_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.17.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.17.mlp.gate_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.17.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.17.mlp.up_proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.17.post_attention_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.18.input_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 32 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.18.linear_attn.A_log" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 8192, + 1, + 4 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.18.linear_attn.conv1d.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 32 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.18.linear_attn.dt_bias" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 32, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.18.linear_attn.in_proj_a.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 32, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.18.linear_attn.in_proj_b.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 8192, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.18.linear_attn.in_proj_qkv.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 4096, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.18.linear_attn.in_proj_z.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 128 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.18.linear_attn.norm.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 4096, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.18.linear_attn.out_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": null, + "shape": [ + 4096, + 12288 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.18.mlp.down_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.18.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.18.mlp.gate_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.18.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.18.mlp.up_proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.18.post_attention_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.19.input_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": null, + "shape": [ + 4096, + 12288 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.19.mlp.down_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.19.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.19.mlp.gate_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.19.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.19.mlp.up_proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.19.post_attention_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 256 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.19.self_attn.k_norm.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": "model.language_model.layers.19.self_attn.qkv_proj", + "shape": [ + 1024, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.19.self_attn.k_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 4096, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.19.self_attn.o_proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 256 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.19.self_attn.q_norm.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": "model.language_model.layers.19.self_attn.qkv_proj", + "shape": [ + 8192, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.19.self_attn.q_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": "model.language_model.layers.19.self_attn.qkv_proj", + "shape": [ + 1024, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.19.self_attn.v_proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00001-of-00004.safetensors", + "tensor_name": "model.language_model.layers.2.input_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 32 + ], + "source_file": "model-00001-of-00004.safetensors", + "tensor_name": "model.language_model.layers.2.linear_attn.A_log" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 8192, + 1, + 4 + ], + "source_file": "model-00001-of-00004.safetensors", + "tensor_name": "model.language_model.layers.2.linear_attn.conv1d.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 32 + ], + "source_file": "model-00001-of-00004.safetensors", + "tensor_name": "model.language_model.layers.2.linear_attn.dt_bias" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 32, + 4096 + ], + "source_file": "model-00001-of-00004.safetensors", + "tensor_name": "model.language_model.layers.2.linear_attn.in_proj_a.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 32, + 4096 + ], + "source_file": "model-00001-of-00004.safetensors", + "tensor_name": "model.language_model.layers.2.linear_attn.in_proj_b.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 8192, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.2.linear_attn.in_proj_qkv.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 4096, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.2.linear_attn.in_proj_z.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 128 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.2.linear_attn.norm.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 4096, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.2.linear_attn.out_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": null, + "shape": [ + 4096, + 12288 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.2.mlp.down_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.2.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.2.mlp.gate_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.2.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.2.mlp.up_proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.2.post_attention_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.20.input_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 32 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.20.linear_attn.A_log" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 8192, + 1, + 4 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.20.linear_attn.conv1d.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 32 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.20.linear_attn.dt_bias" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 32, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.20.linear_attn.in_proj_a.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 32, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.20.linear_attn.in_proj_b.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 8192, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.20.linear_attn.in_proj_qkv.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 4096, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.20.linear_attn.in_proj_z.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 128 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.20.linear_attn.norm.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 4096, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.20.linear_attn.out_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": null, + "shape": [ + 4096, + 12288 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.20.mlp.down_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.20.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.20.mlp.gate_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.20.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.20.mlp.up_proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.20.post_attention_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.21.input_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 32 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.21.linear_attn.A_log" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 8192, + 1, + 4 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.21.linear_attn.conv1d.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 32 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.21.linear_attn.dt_bias" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 32, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.21.linear_attn.in_proj_a.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 32, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.21.linear_attn.in_proj_b.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 8192, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.21.linear_attn.in_proj_qkv.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 4096, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.21.linear_attn.in_proj_z.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 128 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.21.linear_attn.norm.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 4096, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.21.linear_attn.out_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": null, + "shape": [ + 4096, + 12288 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.21.mlp.down_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.21.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.21.mlp.gate_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.21.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.21.mlp.up_proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.21.post_attention_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.22.input_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 32 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.22.linear_attn.A_log" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 8192, + 1, + 4 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.22.linear_attn.conv1d.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 32 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.22.linear_attn.dt_bias" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 32, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.22.linear_attn.in_proj_a.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 32, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.22.linear_attn.in_proj_b.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 8192, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.22.linear_attn.in_proj_qkv.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 4096, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.22.linear_attn.in_proj_z.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 128 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.22.linear_attn.norm.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 4096, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.22.linear_attn.out_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": null, + "shape": [ + 4096, + 12288 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.22.mlp.down_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.22.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.22.mlp.gate_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.22.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.22.mlp.up_proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.22.post_attention_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.23.input_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": null, + "shape": [ + 4096, + 12288 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.23.mlp.down_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.23.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.23.mlp.gate_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.23.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.23.mlp.up_proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.23.post_attention_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 256 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.23.self_attn.k_norm.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": "model.language_model.layers.23.self_attn.qkv_proj", + "shape": [ + 1024, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.23.self_attn.k_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 4096, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.23.self_attn.o_proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 256 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.23.self_attn.q_norm.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": "model.language_model.layers.23.self_attn.qkv_proj", + "shape": [ + 8192, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.23.self_attn.q_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": "model.language_model.layers.23.self_attn.qkv_proj", + "shape": [ + 1024, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.23.self_attn.v_proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.24.input_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 32 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.24.linear_attn.A_log" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 8192, + 1, + 4 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.24.linear_attn.conv1d.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 32 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.24.linear_attn.dt_bias" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 32, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.24.linear_attn.in_proj_a.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 32, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.24.linear_attn.in_proj_b.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 8192, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.24.linear_attn.in_proj_qkv.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 4096, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.24.linear_attn.in_proj_z.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 128 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.24.linear_attn.norm.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 4096, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.24.linear_attn.out_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": null, + "shape": [ + 4096, + 12288 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.24.mlp.down_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.24.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.24.mlp.gate_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.24.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.24.mlp.up_proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.24.post_attention_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.25.input_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 32 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.25.linear_attn.A_log" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 8192, + 1, + 4 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.25.linear_attn.conv1d.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 32 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.25.linear_attn.dt_bias" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 32, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.25.linear_attn.in_proj_a.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 32, + 4096 + ], + "source_file": "model-00003-of-00004.safetensors", + "tensor_name": "model.language_model.layers.25.linear_attn.in_proj_b.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 8192, + 4096 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.25.linear_attn.in_proj_qkv.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 4096, + 4096 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.25.linear_attn.in_proj_z.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 128 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.25.linear_attn.norm.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 4096, + 4096 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.25.linear_attn.out_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": null, + "shape": [ + 4096, + 12288 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.25.mlp.down_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.25.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.25.mlp.gate_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.25.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.25.mlp.up_proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.25.post_attention_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.26.input_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 32 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.26.linear_attn.A_log" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 8192, + 1, + 4 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.26.linear_attn.conv1d.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 32 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.26.linear_attn.dt_bias" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 32, + 4096 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.26.linear_attn.in_proj_a.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 32, + 4096 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.26.linear_attn.in_proj_b.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 8192, + 4096 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.26.linear_attn.in_proj_qkv.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 4096, + 4096 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.26.linear_attn.in_proj_z.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 128 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.26.linear_attn.norm.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 4096, + 4096 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.26.linear_attn.out_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": null, + "shape": [ + 4096, + 12288 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.26.mlp.down_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.26.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.26.mlp.gate_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.26.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.26.mlp.up_proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.26.post_attention_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.27.input_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": null, + "shape": [ + 4096, + 12288 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.27.mlp.down_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.27.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.27.mlp.gate_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.27.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.27.mlp.up_proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.27.post_attention_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 256 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.27.self_attn.k_norm.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": "model.language_model.layers.27.self_attn.qkv_proj", + "shape": [ + 1024, + 4096 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.27.self_attn.k_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 4096, + 4096 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.27.self_attn.o_proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 256 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.27.self_attn.q_norm.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": "model.language_model.layers.27.self_attn.qkv_proj", + "shape": [ + 8192, + 4096 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.27.self_attn.q_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": "model.language_model.layers.27.self_attn.qkv_proj", + "shape": [ + 1024, + 4096 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.27.self_attn.v_proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.28.input_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 32 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.28.linear_attn.A_log" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 8192, + 1, + 4 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.28.linear_attn.conv1d.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 32 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.28.linear_attn.dt_bias" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 32, + 4096 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.28.linear_attn.in_proj_a.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 32, + 4096 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.28.linear_attn.in_proj_b.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 8192, + 4096 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.28.linear_attn.in_proj_qkv.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 4096, + 4096 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.28.linear_attn.in_proj_z.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 128 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.28.linear_attn.norm.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 4096, + 4096 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.28.linear_attn.out_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": null, + "shape": [ + 4096, + 12288 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.28.mlp.down_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.28.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.28.mlp.gate_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.28.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.28.mlp.up_proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.28.post_attention_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.29.input_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 32 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.29.linear_attn.A_log" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 8192, + 1, + 4 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.29.linear_attn.conv1d.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 32 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.29.linear_attn.dt_bias" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 32, + 4096 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.29.linear_attn.in_proj_a.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 32, + 4096 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.29.linear_attn.in_proj_b.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 8192, + 4096 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.29.linear_attn.in_proj_qkv.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 4096, + 4096 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.29.linear_attn.in_proj_z.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 128 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.29.linear_attn.norm.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 4096, + 4096 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.29.linear_attn.out_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": null, + "shape": [ + 4096, + 12288 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.29.mlp.down_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.29.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.29.mlp.gate_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.29.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.29.mlp.up_proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.29.post_attention_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.3.input_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": null, + "shape": [ + 4096, + 12288 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.3.mlp.down_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.3.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.3.mlp.gate_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.3.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.3.mlp.up_proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.3.post_attention_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 256 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.3.self_attn.k_norm.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": "model.language_model.layers.3.self_attn.qkv_proj", + "shape": [ + 1024, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.3.self_attn.k_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 4096, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.3.self_attn.o_proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 256 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.3.self_attn.q_norm.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": "model.language_model.layers.3.self_attn.qkv_proj", + "shape": [ + 8192, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.3.self_attn.q_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": "model.language_model.layers.3.self_attn.qkv_proj", + "shape": [ + 1024, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.3.self_attn.v_proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.30.input_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 32 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.30.linear_attn.A_log" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 8192, + 1, + 4 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.30.linear_attn.conv1d.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 32 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.30.linear_attn.dt_bias" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 32, + 4096 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.30.linear_attn.in_proj_a.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 32, + 4096 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.30.linear_attn.in_proj_b.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 8192, + 4096 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.30.linear_attn.in_proj_qkv.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 4096, + 4096 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.30.linear_attn.in_proj_z.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 128 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.30.linear_attn.norm.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 4096, + 4096 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.30.linear_attn.out_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": null, + "shape": [ + 4096, + 12288 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.30.mlp.down_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.30.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.30.mlp.gate_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.30.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.30.mlp.up_proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.30.post_attention_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.31.input_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": null, + "shape": [ + 4096, + 12288 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.31.mlp.down_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.31.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.31.mlp.gate_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.31.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.31.mlp.up_proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.31.post_attention_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 256 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.31.self_attn.k_norm.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": "model.language_model.layers.31.self_attn.qkv_proj", + "shape": [ + 1024, + 4096 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.31.self_attn.k_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 4096, + 4096 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.31.self_attn.o_proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 256 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.31.self_attn.q_norm.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": "model.language_model.layers.31.self_attn.qkv_proj", + "shape": [ + 8192, + 4096 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.31.self_attn.q_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": "model.language_model.layers.31.self_attn.qkv_proj", + "shape": [ + 1024, + 4096 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.layers.31.self_attn.v_proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.4.input_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 32 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.4.linear_attn.A_log" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 8192, + 1, + 4 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.4.linear_attn.conv1d.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 32 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.4.linear_attn.dt_bias" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 32, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.4.linear_attn.in_proj_a.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 32, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.4.linear_attn.in_proj_b.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 8192, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.4.linear_attn.in_proj_qkv.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 4096, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.4.linear_attn.in_proj_z.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 128 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.4.linear_attn.norm.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 4096, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.4.linear_attn.out_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": null, + "shape": [ + 4096, + 12288 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.4.mlp.down_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.4.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.4.mlp.gate_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.4.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.4.mlp.up_proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.4.post_attention_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.5.input_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 32 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.5.linear_attn.A_log" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 8192, + 1, + 4 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.5.linear_attn.conv1d.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 32 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.5.linear_attn.dt_bias" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 32, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.5.linear_attn.in_proj_a.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 32, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.5.linear_attn.in_proj_b.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 8192, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.5.linear_attn.in_proj_qkv.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 4096, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.5.linear_attn.in_proj_z.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 128 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.5.linear_attn.norm.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 4096, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.5.linear_attn.out_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": null, + "shape": [ + 4096, + 12288 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.5.mlp.down_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.5.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.5.mlp.gate_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.5.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.5.mlp.up_proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.5.post_attention_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.6.input_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 32 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.6.linear_attn.A_log" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 8192, + 1, + 4 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.6.linear_attn.conv1d.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 32 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.6.linear_attn.dt_bias" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 32, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.6.linear_attn.in_proj_a.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 32, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.6.linear_attn.in_proj_b.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 8192, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.6.linear_attn.in_proj_qkv.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 4096, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.6.linear_attn.in_proj_z.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 128 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.6.linear_attn.norm.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 4096, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.6.linear_attn.out_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": null, + "shape": [ + 4096, + 12288 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.6.mlp.down_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.6.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.6.mlp.gate_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.6.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.6.mlp.up_proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.6.post_attention_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.7.input_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": null, + "shape": [ + 4096, + 12288 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.7.mlp.down_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.7.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.7.mlp.gate_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.7.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.7.mlp.up_proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.7.post_attention_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 256 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.7.self_attn.k_norm.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": "model.language_model.layers.7.self_attn.qkv_proj", + "shape": [ + 1024, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.7.self_attn.k_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 4096, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.7.self_attn.o_proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 256 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.7.self_attn.q_norm.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": "model.language_model.layers.7.self_attn.qkv_proj", + "shape": [ + 8192, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.7.self_attn.q_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": "model.language_model.layers.7.self_attn.qkv_proj", + "shape": [ + 1024, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.7.self_attn.v_proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.8.input_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 32 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.8.linear_attn.A_log" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 8192, + 1, + 4 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.8.linear_attn.conv1d.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 32 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.8.linear_attn.dt_bias" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 32, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.8.linear_attn.in_proj_a.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 32, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.8.linear_attn.in_proj_b.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 8192, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.8.linear_attn.in_proj_qkv.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 4096, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.8.linear_attn.in_proj_z.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 128 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.8.linear_attn.norm.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 4096, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.8.linear_attn.out_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": null, + "shape": [ + 4096, + 12288 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.8.mlp.down_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.8.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.8.mlp.gate_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.8.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.8.mlp.up_proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.8.post_attention_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.9.input_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 32 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.9.linear_attn.A_log" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 8192, + 1, + 4 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.9.linear_attn.conv1d.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "attention", + "scale_group": null, + "shape": [ + 32 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.9.linear_attn.dt_bias" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 32, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.9.linear_attn.in_proj_a.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 32, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.9.linear_attn.in_proj_b.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 8192, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.9.linear_attn.in_proj_qkv.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 4096, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.9.linear_attn.in_proj_z.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 128 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.9.linear_attn.norm.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "attention", + "scale_group": null, + "shape": [ + 4096, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.9.linear_attn.out_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": null, + "shape": [ + 4096, + 12288 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.9.mlp.down_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.9.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.9.mlp.gate_proj.weight" + }, + { + "dtype": "BF16", + "method": "nvfp4", + "reason": "Unmeasured native NVFP4 RTN, block 16, activations preserved", + "role": "mlp", + "scale_group": "model.language_model.layers.9.mlp.gate_up_proj", + "shape": [ + 12288, + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.9.mlp.up_proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00002-of-00004.safetensors", + "tensor_name": "model.language_model.layers.9.post_attention_layernorm.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "norm", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.language_model.norm.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.0.attn.proj.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.0.attn.proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 3456 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.0.attn.qkv.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 3456, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.0.attn.qkv.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 4304 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.0.mlp.linear_fc1.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 4304, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.0.mlp.linear_fc1.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.0.mlp.linear_fc2.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152, + 4304 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.0.mlp.linear_fc2.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.0.norm1.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.0.norm1.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.0.norm2.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.0.norm2.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.1.attn.proj.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.1.attn.proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 3456 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.1.attn.qkv.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 3456, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.1.attn.qkv.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 4304 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.1.mlp.linear_fc1.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 4304, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.1.mlp.linear_fc1.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.1.mlp.linear_fc2.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152, + 4304 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.1.mlp.linear_fc2.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.1.norm1.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.1.norm1.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.1.norm2.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.1.norm2.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.10.attn.proj.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.10.attn.proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 3456 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.10.attn.qkv.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 3456, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.10.attn.qkv.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 4304 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.10.mlp.linear_fc1.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 4304, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.10.mlp.linear_fc1.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.10.mlp.linear_fc2.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152, + 4304 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.10.mlp.linear_fc2.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.10.norm1.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.10.norm1.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.10.norm2.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.10.norm2.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.11.attn.proj.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.11.attn.proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 3456 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.11.attn.qkv.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 3456, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.11.attn.qkv.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 4304 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.11.mlp.linear_fc1.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 4304, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.11.mlp.linear_fc1.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.11.mlp.linear_fc2.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152, + 4304 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.11.mlp.linear_fc2.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.11.norm1.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.11.norm1.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.11.norm2.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.11.norm2.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.12.attn.proj.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.12.attn.proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 3456 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.12.attn.qkv.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 3456, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.12.attn.qkv.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 4304 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.12.mlp.linear_fc1.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 4304, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.12.mlp.linear_fc1.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.12.mlp.linear_fc2.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152, + 4304 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.12.mlp.linear_fc2.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.12.norm1.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.12.norm1.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.12.norm2.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.12.norm2.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.13.attn.proj.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.13.attn.proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 3456 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.13.attn.qkv.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 3456, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.13.attn.qkv.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 4304 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.13.mlp.linear_fc1.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 4304, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.13.mlp.linear_fc1.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.13.mlp.linear_fc2.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152, + 4304 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.13.mlp.linear_fc2.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.13.norm1.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.13.norm1.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.13.norm2.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.13.norm2.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.14.attn.proj.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.14.attn.proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 3456 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.14.attn.qkv.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 3456, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.14.attn.qkv.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 4304 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.14.mlp.linear_fc1.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 4304, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.14.mlp.linear_fc1.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.14.mlp.linear_fc2.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152, + 4304 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.14.mlp.linear_fc2.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.14.norm1.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.14.norm1.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.14.norm2.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.14.norm2.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.15.attn.proj.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.15.attn.proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 3456 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.15.attn.qkv.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 3456, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.15.attn.qkv.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 4304 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.15.mlp.linear_fc1.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 4304, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.15.mlp.linear_fc1.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.15.mlp.linear_fc2.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152, + 4304 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.15.mlp.linear_fc2.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.15.norm1.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.15.norm1.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.15.norm2.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.15.norm2.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.16.attn.proj.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.16.attn.proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 3456 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.16.attn.qkv.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 3456, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.16.attn.qkv.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 4304 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.16.mlp.linear_fc1.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 4304, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.16.mlp.linear_fc1.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.16.mlp.linear_fc2.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152, + 4304 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.16.mlp.linear_fc2.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.16.norm1.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.16.norm1.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.16.norm2.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.16.norm2.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.17.attn.proj.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.17.attn.proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 3456 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.17.attn.qkv.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 3456, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.17.attn.qkv.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 4304 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.17.mlp.linear_fc1.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 4304, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.17.mlp.linear_fc1.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.17.mlp.linear_fc2.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152, + 4304 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.17.mlp.linear_fc2.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.17.norm1.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.17.norm1.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.17.norm2.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.17.norm2.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.18.attn.proj.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.18.attn.proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 3456 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.18.attn.qkv.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 3456, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.18.attn.qkv.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 4304 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.18.mlp.linear_fc1.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 4304, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.18.mlp.linear_fc1.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.18.mlp.linear_fc2.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152, + 4304 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.18.mlp.linear_fc2.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.18.norm1.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.18.norm1.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.18.norm2.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.18.norm2.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.19.attn.proj.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.19.attn.proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 3456 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.19.attn.qkv.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 3456, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.19.attn.qkv.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 4304 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.19.mlp.linear_fc1.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 4304, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.19.mlp.linear_fc1.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.19.mlp.linear_fc2.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152, + 4304 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.19.mlp.linear_fc2.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.19.norm1.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.19.norm1.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.19.norm2.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.19.norm2.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.2.attn.proj.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.2.attn.proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 3456 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.2.attn.qkv.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 3456, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.2.attn.qkv.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 4304 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.2.mlp.linear_fc1.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 4304, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.2.mlp.linear_fc1.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.2.mlp.linear_fc2.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152, + 4304 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.2.mlp.linear_fc2.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.2.norm1.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.2.norm1.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.2.norm2.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.2.norm2.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.20.attn.proj.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.20.attn.proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 3456 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.20.attn.qkv.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 3456, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.20.attn.qkv.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 4304 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.20.mlp.linear_fc1.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 4304, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.20.mlp.linear_fc1.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.20.mlp.linear_fc2.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152, + 4304 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.20.mlp.linear_fc2.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.20.norm1.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.20.norm1.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.20.norm2.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.20.norm2.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.21.attn.proj.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.21.attn.proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 3456 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.21.attn.qkv.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 3456, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.21.attn.qkv.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 4304 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.21.mlp.linear_fc1.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 4304, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.21.mlp.linear_fc1.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.21.mlp.linear_fc2.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152, + 4304 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.21.mlp.linear_fc2.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.21.norm1.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.21.norm1.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.21.norm2.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.21.norm2.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.22.attn.proj.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.22.attn.proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 3456 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.22.attn.qkv.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 3456, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.22.attn.qkv.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 4304 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.22.mlp.linear_fc1.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 4304, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.22.mlp.linear_fc1.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.22.mlp.linear_fc2.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152, + 4304 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.22.mlp.linear_fc2.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.22.norm1.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.22.norm1.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.22.norm2.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.22.norm2.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.23.attn.proj.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.23.attn.proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 3456 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.23.attn.qkv.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 3456, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.23.attn.qkv.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 4304 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.23.mlp.linear_fc1.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 4304, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.23.mlp.linear_fc1.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.23.mlp.linear_fc2.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152, + 4304 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.23.mlp.linear_fc2.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.23.norm1.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.23.norm1.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.23.norm2.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.23.norm2.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.24.attn.proj.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.24.attn.proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 3456 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.24.attn.qkv.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 3456, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.24.attn.qkv.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 4304 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.24.mlp.linear_fc1.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 4304, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.24.mlp.linear_fc1.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.24.mlp.linear_fc2.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152, + 4304 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.24.mlp.linear_fc2.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.24.norm1.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.24.norm1.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.24.norm2.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.24.norm2.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.25.attn.proj.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.25.attn.proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 3456 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.25.attn.qkv.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 3456, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.25.attn.qkv.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 4304 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.25.mlp.linear_fc1.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 4304, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.25.mlp.linear_fc1.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.25.mlp.linear_fc2.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152, + 4304 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.25.mlp.linear_fc2.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.25.norm1.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.25.norm1.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.25.norm2.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.25.norm2.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.26.attn.proj.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.26.attn.proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 3456 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.26.attn.qkv.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 3456, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.26.attn.qkv.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 4304 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.26.mlp.linear_fc1.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 4304, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.26.mlp.linear_fc1.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.26.mlp.linear_fc2.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152, + 4304 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.26.mlp.linear_fc2.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.26.norm1.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.26.norm1.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.26.norm2.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.26.norm2.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.3.attn.proj.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.3.attn.proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 3456 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.3.attn.qkv.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 3456, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.3.attn.qkv.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 4304 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.3.mlp.linear_fc1.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 4304, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.3.mlp.linear_fc1.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.3.mlp.linear_fc2.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152, + 4304 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.3.mlp.linear_fc2.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.3.norm1.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.3.norm1.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.3.norm2.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.3.norm2.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.4.attn.proj.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.4.attn.proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 3456 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.4.attn.qkv.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 3456, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.4.attn.qkv.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 4304 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.4.mlp.linear_fc1.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 4304, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.4.mlp.linear_fc1.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.4.mlp.linear_fc2.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152, + 4304 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.4.mlp.linear_fc2.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.4.norm1.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.4.norm1.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.4.norm2.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.4.norm2.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.5.attn.proj.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.5.attn.proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 3456 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.5.attn.qkv.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 3456, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.5.attn.qkv.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 4304 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.5.mlp.linear_fc1.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 4304, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.5.mlp.linear_fc1.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.5.mlp.linear_fc2.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152, + 4304 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.5.mlp.linear_fc2.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.5.norm1.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.5.norm1.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.5.norm2.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.5.norm2.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.6.attn.proj.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.6.attn.proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 3456 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.6.attn.qkv.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 3456, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.6.attn.qkv.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 4304 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.6.mlp.linear_fc1.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 4304, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.6.mlp.linear_fc1.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.6.mlp.linear_fc2.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152, + 4304 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.6.mlp.linear_fc2.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.6.norm1.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.6.norm1.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.6.norm2.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.6.norm2.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.7.attn.proj.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.7.attn.proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 3456 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.7.attn.qkv.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 3456, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.7.attn.qkv.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 4304 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.7.mlp.linear_fc1.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 4304, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.7.mlp.linear_fc1.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.7.mlp.linear_fc2.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152, + 4304 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.7.mlp.linear_fc2.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.7.norm1.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.7.norm1.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.7.norm2.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.7.norm2.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.8.attn.proj.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.8.attn.proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 3456 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.8.attn.qkv.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 3456, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.8.attn.qkv.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 4304 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.8.mlp.linear_fc1.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 4304, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.8.mlp.linear_fc1.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.8.mlp.linear_fc2.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152, + 4304 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.8.mlp.linear_fc2.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.8.norm1.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.8.norm1.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.8.norm2.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.8.norm2.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.9.attn.proj.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.9.attn.proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 3456 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.9.attn.qkv.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 3456, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.9.attn.qkv.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 4304 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.9.mlp.linear_fc1.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 4304, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.9.mlp.linear_fc1.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.9.mlp.linear_fc2.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152, + 4304 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.9.mlp.linear_fc2.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.9.norm1.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.9.norm1.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.9.norm2.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.blocks.9.norm2.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 4608 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.merger.linear_fc1.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 4608, + 4608 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.merger.linear_fc1.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 4096 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.merger.linear_fc2.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 4096, + 4608 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.merger.linear_fc2.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.merger.norm.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.merger.norm.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.patch_embed.proj.bias" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 1152, + 3, + 2, + 16, + 16 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.patch_embed.proj.weight" + }, + { + "dtype": "BF16", + "method": "preserve", + "reason": "Source precision preserved by protection, shape, dtype or keep policy", + "role": "vision", + "scale_group": null, + "shape": [ + 2304, + 1152 + ], + "source_file": "model-00004-of-00004.safetensors", + "tensor_name": "model.visual.pos_embed.weight" + } + ], + "block_size": 16, + "config_sha256": "466925e846dede7658a49cddc47256a9bbd9056445e91b088e9f490fad3a956e", + "created_at": "2026-10-11T14:07:19.016306Z", + "estimated_weight_bytes": 8874278336, + "evidence_kind": "unmeasured", + "format": "nvfp4", + "keep_patterns": [], + "model_id": "Cloudflare/clef-flash", + "model_type": "qwen3_5", + "revision": "8b2e5fd17c09fd49bc2880b805d5323ec7ff4ff3", + "schema_version": "axquant.cuda-plan.v1", + "source_files": [ + { + "path": "chat_template.jinja", + "sha256": "a4aee8afcf2e0711942cf848899be66016f8d14a889ff9ede07bca099c28f715", + "size_bytes": 7756 + }, + { + "path": "config.json", + "sha256": "66f87f6fb2616b46604daf2a9c67ddc87938296d07156efa34d59b5be49e3238", + "size_bytes": 2832 + }, + { + "path": "generation_config.json", + "sha256": "45707f8467bf4e54e112c6095a1b2d1f2d63651cf86816344cfff21f02dda0d4", + "size_bytes": 116 + }, + { + "path": "joint_schema_model.py", + "sha256": "0e304cf7c6500e8bb59bef7e2afd2c6373f82596dfb3b57d1aa93c175e2dc3a3", + "size_bytes": 23263 + }, + { + "path": "model-00001-of-00004.safetensors", + "sha256": "8b45a8e968141cdcc58fb71c9adfc258e2c77b5f062bc636c1fd5bc5d916b565", + "size_bytes": 4942706120 + }, + { + "path": "model-00002-of-00004.safetensors", + "sha256": "7590856c713eed844a2dcf48e6c43c4de165b788bc3f80e328311183cdbc7db8", + "size_bytes": 4987757928 + }, + { + "path": "model-00003-of-00004.safetensors", + "sha256": "e6eac2467952c33361ed7dcb3c7959d1086bbe57201cd3749c3d769fdc17fe63", + "size_bytes": 4954810240 + }, + { + "path": "model-00004-of-00004.safetensors", + "sha256": "9fcecc6556b39171238373a465f409794b7f821fb4cd1e6459e3a9c0fe317af7", + "size_bytes": 3934446832 + }, + { + "path": "model.safetensors.index.json", + "sha256": "941305ff9f77551e145a6cea976ef456cd5cb208cbece99f168376c752fcf96c", + "size_bytes": 69253 + }, + { + "path": "processor_config.json", + "sha256": "d89ef49ce9cd37fbf510158e13c1ef063d9286411c1ec9049932dbe0487143b1", + "size_bytes": 1191 + }, + { + "path": "tokenizer.json", + "sha256": "06b9509352d2af50381ab2247e083b80d32d5c0aba91c272ca9ff729b6a0e523", + "size_bytes": 19989325 + }, + { + "path": "tokenizer_config.json", + "sha256": "91a08f825d370d085d692e04cf117cdd7faad7bf18e996f1e6031b6dab03db72", + "size_bytes": 1075 + } + ] +}