ColorShadow-7B-v3
This is a Gradient-SLERP merge between ColorShadow-7B and Terminis-7B performed using mergekit.
Here is the config file used:
slices:
- sources:
- model: nlpguy/ColorShadow-7B
layer_range: [0, 32]
- model: Q-bert/Terminis-7B
layer_range: [0, 32]
merge_method: slerp
base_model: nlpguy/ColorShadow-7B
parameters:
t:
- filter: self_attn
value: [1, 0.5, 0.7, 0.3, 0]
- filter: mlp
value: [0, 0.5, 0.3, 0.7, 1]
- value: 0.5 # fallback for rest of tensors
dtype: float16
Open LLM Leaderboard Evaluation Results
Detailed results can be found here
| Metric | Value |
|---|---|
| Avg. | 67.29 |
| AI2 Reasoning Challenge (25-Shot) | 67.58 |
| HellaSwag (10-Shot) | 85.04 |
| MMLU (5-Shot) | 60.57 |
| TruthfulQA (0-shot) | 62.88 |
| Winogrande (5-shot) | 80.11 |
| GSM8k (5-shot) | 47.54 |
- Downloads last month
- 1,600
Inference Providers
NEW
This model is not currently available via any of the supported third-party Inference Providers, and
the model is not deployed on the HF Inference API.
Model tree for nlpguy/ColorShadow-7B-v3
Evaluation results
- normalized accuracy on AI2 Reasoning Challenge (25-Shot)test set Open LLM Leaderboard67.580
- normalized accuracy on HellaSwag (10-Shot)validation set Open LLM Leaderboard85.040
- accuracy on MMLU (5-Shot)test set Open LLM Leaderboard60.570
- mc2 on TruthfulQA (0-shot)validation set Open LLM Leaderboard62.880
- accuracy on Winogrande (5-shot)validation set Open LLM Leaderboard80.110
- accuracy on GSM8k (5-shot)test set Open LLM Leaderboard47.540