1.3 KiB
1.3 KiB
base_model, tags, license, language
| base_model | tags | license | language | |||||
|---|---|---|---|---|---|---|---|---|
| unsloth/Qwen3-0.6B-unsloth-bnb-4bit |
|
apache-2.0 |
|
Meet7 0.6B
A general-purpose Non-Reasoning LoRA fine-tune of Qwen3-0.6B, trained in under 10 minutes on just 600 samples.
Benchmarks
Scores are acc_norm.
| Task | Shot | Qwen3-0.6B | Meet7 0.6B | Δ |
|---|---|---|---|---|
| BoolQ | 0-shot | 0.3798 | 0.5554 | +17.56% |
| ARC Easy | 3-shot | 0.3636 | 0.4394 | +07.58% |
| ARC Challenge | 3-shot | 0.2952 | 0.3456 | +05.04% |
| HellaSwag | 3-shot | 0.3956 | 0.4323 | +03.67% |
| PIQA | 0-shot | 0.6338 | 0.6583 | +02.45% |
| Winogrande | 0-shot | 0.5225 | 0.5201 | −00.24% |
Model Details
| Developed by | Ma7ee7 |
| License | Apache-2.0 |
| Base model | unsloth/Qwen3-0.6B-unsloth-bnb-4bit |
| Training time | < 10 minutes |
| Training samples | 600 |
Trained 2x faster with Unsloth and Hugging Face TRL.
