Model soups: averaging weights of multiple fine-tuned models improves accuracy without increasing inference time
Paper
•
2203.05482
•
Published
•
7
merge
This is a merge of pre-trained language models created using mergekit.
Merge Details
Merge Method
This model was merged using the Linear merge method.
Models Merged
The following models were included in the merge:
- DreadPoor/Rusted_Platinum-8B-LINEAR
- DreadPoor/H_the_eighth-8B-LINEAR
- DreadPoor/Spring_Dusk-8B-SCE
- DreadPoor/BaeZel-8B-LINEAR
Configuration
The following YAML configuration was used to produce this model:
models:
- model: DreadPoor/H_the_eighth-8B-LINEAR
parameters:
weight: 1.0
- model: DreadPoor/Spring_Dusk-8B-SCE
parameters:
weight: 1.0
- model: DreadPoor/Rusted_Platinum-8B-LINEAR
parameters:
weight: 1.0
- model: DreadPoor/BaeZel-8B-LINEAR
parameters:
weight: 1.0
merge_method: linear
normalize: true
int8_mask: true
dtype: bfloat16
Open LLM Leaderboard Evaluation Results
Detailed results can be found here! Summarized results can be found here!
| Metric | Value (%) |
|---|---|
| Average | 29.61 |
| IFEval (0-Shot) | 75.27 |
| BBH (3-Shot) | 36.20 |
| MATH Lvl 5 (4-Shot) | 17.30 |
| GPQA (0-shot) | 6.38 |
| MuSR (0-shot) | 11.10 |
| MMLU-PRO (5-shot) | 31.41 |
- Downloads last month
- 15
Model tree for DreadPoor/inexpertus_1.1-8B-LINEAR
Merge model
this model
Paper for DreadPoor/inexpertus_1.1-8B-LINEAR
Evaluation results
- averaged accuracy on IFEval (0-Shot)Open LLM Leaderboard75.270
- normalized accuracy on BBH (3-Shot)test set Open LLM Leaderboard36.200
- exact match on MATH Lvl 5 (4-Shot)test set Open LLM Leaderboard17.300
- acc_norm on GPQA (0-shot)Open LLM Leaderboard6.380
- acc_norm on MuSR (0-shot)Open LLM Leaderboard11.100
- accuracy on MMLU-PRO (5-shot)test set Open LLM Leaderboard31.410