|
|
--- |
|
|
library_name: transformers |
|
|
tags: |
|
|
- small-lm |
|
|
- math |
|
|
- code |
|
|
- reasoning |
|
|
- slm |
|
|
- french |
|
|
- merge |
|
|
license: apache-2.0 |
|
|
base_model: |
|
|
- SamsungSAILMontreal/Qwen3-1.7B |
|
|
- SamsungSAILMontreal/Qwen3-1.7B-Math |
|
|
- SamsungSAILMontreal/Qwen3-1.7B-Code |
|
|
- SamsungSAILMontreal/Qwen3-1.7B-Fr |
|
|
language: |
|
|
- fr |
|
|
- en |
|
|
--- |
|
|
|
|
|
# Qwen3-1.7B-Math-Fr |
|
|
|
|
|
This model is obtained by merging [SamsungSAILMontreal/Qwen3-1.7B-Math](https://huggingface.co/SamsungSAILMontreal/Qwen3-1.7B-Math), |
|
|
[SamsungSAILMontreal/Qwen3-1.7B-Code](https://huggingface.co/SamsungSAILMontreal/Qwen3-1.7B-Code) and |
|
|
[SamsungSAILMontreal/Qwen3-1.7B-Fr](https://huggingface.co/SamsungSAILMontreal/Qwen3-1.7B-Fr). |
|
|
The model is used in the experiments described in https://bknyaz.github.io/blog/2026/meta-merge/. |
|
|
Single A100 was used for merging and evaluation. |
|
|
|
|
|
The following versions were used for merge/eval: |
|
|
|
|
|
- python >= 3.10 |
|
|
- torch : 2.9.0+cu128 |
|
|
- lm_eval : 0.4.9.1 |
|
|
- vllm : 0.11.1 |
|
|
- transformers : 4.57.6 |
|
|
- datasets : 3.2.0 |
|
|
- numpy : 2.2.6 |
|
|
|
|
|
## Merging |
|
|
|
|
|
Merging was done using parameter averaging implemented in [merge_qwen.py](https://github.com/SamsungSAILMontreal/nino/blob/main/merge_qwen.py). |
|
|
|
|
|
## Evaluation |
|
|
|
|
|
Evaluation was done with lm_eval on the test split of [gsm8k](https://huggingface.co/datasets/openai/gsm8k), |
|
|
[french_bench (avg score)](https://github.com/EleutherAI/lm-evaluation-harness/tree/main/lm_eval/tasks/french_bench), |
|
|
[gsm8k-fr](https://huggingface.co/datasets/cmh/gsm8k_fr) and humaneval (instruct): |
|
|
|
|
|
```bash |
|
|
python -m lm_eval --model vllm --model_args pretrained=${model},tensor_parallel_size=1,dtype=auto,gpu_memory_utilization=0.9,data_parallel_size=1 \ |
|
|
--tasks gsm8k,french_bench,gsm8k-fr,humaneval_instruct --batch_size 1 --apply_chat_template=True --confirm_run_unsafe_code --trust_remote_code |
|
|
``` |
|
|
|
|
|
To evaluate on gsm8k-fr you can use our fork https://github.com/bknyaz/lm-evaluation-harness/tree/main/lm_eval/tasks/gsm8k. |
|
|
|
|
|
### Results |
|
|
|
|
|
| Model | gsm8k | french | gsm8k-fr | humaneval_instruct | avg | |
|
|
|---------------------------|-------|--------|----------|--------------------|------| |
|
|
| Qwen3-1.7B | 20.6 | 26.2 | 20.2 | 67.1 | 33.5 | |
|
|
| Qwen3-1.7B-Math | 62.1 | 28.3 | 41.5 | 48.2 | 45.0 | |
|
|
| Qwen3-1.7B-Fr | 60.9 | 32.8 | 43.9 | 56.1 | 48.4 | |
|
|
| Qwen3-1.7B-Code | 56.7 | 28.2 | 36.5 | 69.5 | 47.7 | |
|
|
| Qwen3-1.7B-Math-Code-Fr | 64.6 | 29.2 | 48.4 | 65.2 | 51.8 | |
|
|
|
|
|
## License |
|
|
|
|
|
Please refer to the license of the base models [SamsungSAILMontreal/Qwen3-1.7B-Math](https://huggingface.co/SamsungSAILMontreal/Qwen3-1.7B-Math), |
|
|
[SamsungSAILMontreal/Qwen3-1.7B-Code](https://huggingface.co/SamsungSAILMontreal/Qwen3-1.7B-Code) and |
|
|
[SamsungSAILMontreal/Qwen3-1.7B-Fr](https://huggingface.co/SamsungSAILMontreal/Qwen3-1.7B-Fr). |
