初始化项目，由ModelHub XC社区提供模型
Model: dipta007/GanitLLM-0.6B_SFT_CGRPO Source: Original Platform
This commit is contained in:
ModelHub XC
136
README.md
Normal file
136
README.md
Normal file
@@ -0,0 +1,136 @@
|
||||
---
|
||||
library_name: transformers
|
||||
license: apache-2.0
|
||||
base_model: Qwen/Qwen3-0.6B
|
||||
pipeline_tag: text-generation
|
||||
language:
|
||||
- bn
|
||||
- en
|
||||
tags:
|
||||
- math
|
||||
- bengali
|
||||
- reasoning
|
||||
- grpo
|
||||
- curriculum-learning
|
||||
datasets:
|
||||
- dipta007/Ganit
|
||||
---
|
||||
|
||||
# GanitLLM-0.6B_SFT_CGRPO
|
||||
|
||||
<p align="center">
|
||||
<a href="https://arxiv.org/abs/2601.06767">
|
||||
<img src="https://img.shields.io/badge/%F0%9F%94%A5_Accepted_at-ACL_2026_(Findings)_%F0%9F%94%A5-b12a00?style=for-the-badge&labelColor=ffb300" alt="Accepted at ACL 2026 (Findings)">
|
||||
</a>
|
||||
</p>
|
||||
|
||||
[](https://arxiv.org/abs/2601.06767)
|
||||
[](https://arxiv.org/abs/2601.06767)
|
||||
[](https://dipta007.github.io/GanitLLM/)
|
||||
[](https://huggingface.co/datasets/dipta007/Ganit)
|
||||
[](https://huggingface.co/collections/dipta007/ganitllm)
|
||||
[](https://github.com/dipta007/GanitLLM)
|
||||
|
||||
## Highlights
|
||||
|
||||
**GanitLLM-0.6B_SFT_CGRPO** is our smallest Bengali mathematical reasoning model trained using the novel **Curriculum-GRPO** approach. Ideal for resource-constrained deployments. Key improvements over the base Qwen3-0.6B model:
|
||||
|
||||
- **+20.0 accuracy** on Bn-MGSM benchmark (8.4 → 28.4)
|
||||
- **+40.2 accuracy** on Bn-MSVAMP benchmark (12.2 → 52.4)
|
||||
- **88.50% Bengali reasoning** (vs 12.43% for base model)
|
||||
- **80.4% fewer tokens** in generated solutions (1265 → 248 words)
|
||||
|
||||
## Model Overview
|
||||
|
||||
| Property | Value |
|
||||
|----------|-------|
|
||||
| **Model Type** | Causal Language Model |
|
||||
| **Base Model** | Qwen/Qwen3-0.6B |
|
||||
| **Parameters** | 0.6B |
|
||||
| **Training** | SFT + Curriculum-GRPO |
|
||||
| **Context Length** | 4,096 tokens |
|
||||
| **Language** | Bengali, English |
|
||||
|
||||
## Training Details
|
||||
|
||||
This model was trained using our multi-stage pipeline:
|
||||
|
||||
1. **Supervised Fine-Tuning (SFT)**: Trained on GANIT-SFT (~11k examples) to ground reasoning in Bengali
|
||||
2. **Curriculum-GRPO**: Reinforcement learning with difficulty-aware sampling on GANIT-RLVR (~7.3k examples)
|
||||
|
||||
### Reward Functions
|
||||
- **Format Reward**: Validates `<think>` and `<answer>` tag structure
|
||||
- **Correctness Reward**: +2.0 for Bengali answer match, +1.0 for English match
|
||||
- **Bengali Reasoning Reward**: Ensures >80% Bengali text in reasoning
|
||||
|
||||
## Quickstart
|
||||
|
||||
```python
|
||||
from transformers import AutoModelForCausalLM, AutoTokenizer
|
||||
|
||||
model_name = "dipta007/GanitLLM-0.6B_SFT_CGRPO"
|
||||
|
||||
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
||||
model = AutoModelForCausalLM.from_pretrained(
|
||||
model_name,
|
||||
torch_dtype="auto",
|
||||
device_map="auto"
|
||||
)
|
||||
|
||||
problem = "একটি দোকানে ১২টি আপেল আছে। যদি ৫টি আপেল বিক্রি হয়, তাহলে কতটি আপেল বাকি থাকবে?"
|
||||
|
||||
prompt = f"""A conversation takes place between the user and the assistant. The user asks a question, and the assistant solves the problem. Please reason step by step in Bengali, and put your final answer in the <answer> </answer> tags.
|
||||
|
||||
Question: {problem}"""
|
||||
|
||||
messages = [{"role": "user", "content": prompt}]
|
||||
text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
|
||||
model_inputs = tokenizer([text], return_tensors="pt").to(model.device)
|
||||
|
||||
generated_ids = model.generate(**model_inputs, max_new_tokens=2048, temperature=0.7)
|
||||
output_ids = generated_ids[0][len(model_inputs.input_ids[0]):].tolist()
|
||||
response = tokenizer.decode(output_ids, skip_special_tokens=True)
|
||||
print(response)
|
||||
```
|
||||
|
||||
### Using vLLM
|
||||
|
||||
```bash
|
||||
vllm serve dipta007/GanitLLM-0.6B_SFT_CGRPO --max-model-len 4096
|
||||
```
|
||||
|
||||
## Performance
|
||||
|
||||
| Model | Bn-MGSM | Bn-MSVAMP | Avg. Words | Bengali % |
|
||||
|-------|---------|-----------|------------|-----------|
|
||||
| Qwen3-0.6B (base) | 8.40 | 12.20 | 1265 | 12.43% |
|
||||
| **GanitLLM-0.6B_SFT_CGRPO** | **28.40** | **52.40** | **248** | **88.50%** |
|
||||
|
||||
## Related Models
|
||||
|
||||
| Model | Parameters | Training | Link |
|
||||
|-------|------------|----------|------|
|
||||
| GanitLLM-4B_SFT_CGRPO | 4B | SFT + CGRPO | [Link](https://huggingface.co/dipta007/GanitLLM-4B_SFT_CGRPO) |
|
||||
| GanitLLM-1.7B_SFT_CGRPO | 1.7B | SFT + CGRPO | [Link](https://huggingface.co/dipta007/GanitLLM-1.7B_SFT_CGRPO) |
|
||||
| **GanitLLM-0.6B_SFT_CGRPO** | 0.6B | SFT + CGRPO | [Link](https://huggingface.co/dipta007/GanitLLM-0.6B_SFT_CGRPO) |
|
||||
| GanitLLM-0.6B_SFT_GRPO | 0.6B | SFT + GRPO | [Link](https://huggingface.co/dipta007/GanitLLM-0.6B_SFT_GRPO) |
|
||||
| GanitLLM-0.6B_CGRPO | 0.6B | CGRPO | [Link](https://huggingface.co/dipta007/GanitLLM-0.6B_CGRPO) |
|
||||
|
||||
## Citation
|
||||
|
||||
```bibtex
|
||||
@inproceedings{dipta2026ganitllm,
|
||||
title={GanitLLM: Difficulty-Aware Bengali Mathematical Reasoning through Curriculum-GRPO},
|
||||
author={Shubhashis Roy Dipta and Khairul Mahbub and Nadia Najjar},
|
||||
booktitle={Findings of the Association for Computational Linguistics: ACL 2026},
|
||||
year={2026},
|
||||
eprint={2601.06767},
|
||||
archivePrefix={arXiv},
|
||||
primaryClass={cs.CL},
|
||||
url={https://arxiv.org/abs/2601.06767},
|
||||
}
|
||||
```
|
||||
|
||||
## License
|
||||
|
||||
This model is released under the Apache 2.0 License.
|
||||
Reference in New Issue
Block a user