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:
Configuration
The following YAML configuration was used to produce this model:
models:
- model: DreadPoor/Aspire-8B-model_stock
parameters:
weight: 1.0
- model: DreadPoor/ichor_1.1-8B-Model_Stock
parameters:
weight: 1.0
- model: BoltMonkey/DreadMix
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 | 30.19 |
| IFEval (0-Shot) | 74.69 |
| BBH (3-Shot) | 34.15 |
| MATH Lvl 5 (4-Shot) | 17.75 |
| GPQA (0-shot) | 10.40 |
| MuSR (0-shot) | 12.76 |
| MMLU-PRO (5-shot) | 31.38 |
- Downloads last month
- 6
Model tree for DreadPoor/H_the_eighth-8B-LINEAR
Merge model
this model
Paper for DreadPoor/H_the_eighth-8B-LINEAR
Evaluation results
- averaged accuracy on IFEval (0-Shot)Open LLM Leaderboard74.690
- normalized accuracy on BBH (3-Shot)test set Open LLM Leaderboard34.150
- exact match on MATH Lvl 5 (4-Shot)test set Open LLM Leaderboard17.750
- acc_norm on GPQA (0-shot)Open LLM Leaderboard10.400
- acc_norm on MuSR (0-shot)Open LLM Leaderboard12.760
- accuracy on MMLU-PRO (5-shot)test set Open LLM Leaderboard31.380