Sam Purkis's picture

Sam Purkis

SamPurkis

·

AI & ML interests

None yet

Recent Activity

reacted to eaddario's post with 👍 about 1 month ago

Experimental global target bits‑per‑weight quantization of mistralai/Ministral-3-14B-Instruct-2512 and mistralai/Ministral-3-14B-Reasoning-2512 Unlike standard llama.cpp quantizations that rely on fixed type heuristics (e.g., Q4_K_M), the Target BPW approach optimizes per-tensor precision where it matters the most, and produces high quality models that meet a precise global file size target. Key Advantages: - VRAM Maximization: Can generate high quality models sized exactly to fit hardware constraints (e.g., fitting the model into exactly 24GB VRAM). - Data-Driven Precision: Quantization mix is determined by actual weight error sensitivity rather than hardcoded rules, often yielding better PPL/KLD size trade-offs. Full benchmarks (PPL, KLD, ARC, MMLU, etc.) and methodology in the models' cards https://huggingface.co/eaddario/Ministral-3-14B-Instruct-2512-GGUF https://huggingface.co/eaddario/Ministral-3-14B-Reasoning-2512-GGUF

liked a Space about 2 months ago

published a model 10 months ago

SamPurkis/Qwen3-16B-A1.5B

View all activity

Organizations

None yet

liked a Space about 2 months ago

Qwen3-TTS Demo

Generate speech audio via voice design, cloning, or preset speakers

liked a dataset almost 2 years ago

TIGER-Lab/MMLU-Pro

Benchmark • Updated 8 days ago • 12.1k • 128k • 456

liked a Space almost 2 years ago

Low-bit Quantized Open LLM Leaderboard

Track, rank and evaluate open LLMs and chatbots

liked 2 Spaces over 2 years ago

Open LLM Leaderboard

Track, rank and evaluate open LLMs and chatbots

Open LLM Leaderboard