A Fishy Model
This model was trained with SFT using Unsloth on the ChatML format with 8k context. Carp models are trained with a combination of pretrain, instruct, and chat datasets.
Changes
- Training dataset had some "slop" and refusals removed.
- Datasets were reformatted.
Uploaded model
- Developed by: TheTsar1209
- License: apache-2.0
- Finetuned from model : unsloth/Qwen2.5-14B-Instruct-bnb-4bit
This qwen2 model was trained 2x faster with Unsloth and Huggingface's TRL library.
Open LLM Leaderboard Evaluation Results
Detailed results can be found here
| Metric | Value |
|---|---|
| Avg. | 35.67 |
| IFEval (0-Shot) | 72.02 |
| BBH (3-Shot) | 49.38 |
| MATH Lvl 5 (4-Shot) | 17.37 |
| GPQA (0-shot) | 13.65 |
| MuSR (0-shot) | 15.55 |
| MMLU-PRO (5-shot) | 46.04 |
- Downloads last month
- 36
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social
visibility and check back later, or deploy to Inference Endpoints (dedicated)
instead.
Model tree for TheTsar1209/qwen-carpmuscle-v0.4
Base model
Qwen/Qwen2.5-14B
Finetuned
Qwen/Qwen2.5-14B-Instruct
Quantized
unsloth/Qwen2.5-14B-Instruct-bnb-4bit
Evaluation results
- strict accuracy on IFEval (0-Shot)Open LLM Leaderboard72.020
- normalized accuracy on BBH (3-Shot)Open LLM Leaderboard49.380
- exact match on MATH Lvl 5 (4-Shot)Open LLM Leaderboard17.370
- acc_norm on GPQA (0-shot)Open LLM Leaderboard13.650
- acc_norm on MuSR (0-shot)Open LLM Leaderboard15.550
- accuracy on MMLU-PRO (5-shot)test set Open LLM Leaderboard46.040