Instructions to use dipta007/GanitLLM-4B_SFT_CGRPO with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use dipta007/GanitLLM-4B_SFT_CGRPO with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="dipta007/GanitLLM-4B_SFT_CGRPO")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("dipta007/GanitLLM-4B_SFT_CGRPO")
model = AutoModelForCausalLM.from_pretrained("dipta007/GanitLLM-4B_SFT_CGRPO", device_map="auto")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use dipta007/GanitLLM-4B_SFT_CGRPO with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "dipta007/GanitLLM-4B_SFT_CGRPO"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "dipta007/GanitLLM-4B_SFT_CGRPO",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/dipta007/GanitLLM-4B_SFT_CGRPO

SGLang

How to use dipta007/GanitLLM-4B_SFT_CGRPO with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "dipta007/GanitLLM-4B_SFT_CGRPO" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "dipta007/GanitLLM-4B_SFT_CGRPO",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "dipta007/GanitLLM-4B_SFT_CGRPO" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "dipta007/GanitLLM-4B_SFT_CGRPO",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use dipta007/GanitLLM-4B_SFT_CGRPO with Docker Model Runner:
```
docker model run hf.co/dipta007/GanitLLM-4B_SFT_CGRPO
```

GanitLLM-4B_SFT_CGRPO / README.md

dipta007

Update README.md

54762e3 verified 4 months ago

preview code

Raw

History Blame Contribute Delete

5.17 kB

metadata

library_name: transformers
license: apache-2.0
base_model: Qwen/Qwen3-4B
pipeline_tag: text-generation
language:
  - bn
  - en
tags:
  - math
  - bengali
  - reasoning
  - grpo
  - curriculum-learning
datasets:
  - dipta007/Ganit

GanitLLM-4B_SFT_CGRPO

Highlights

GanitLLM-4B_SFT_CGRPO is our flagship Bengali mathematical reasoning model trained using the novel Curriculum-GRPO approach. Key improvements over the base Qwen3-4B model:

+7.6 accuracy on Bn-MGSM benchmark (69.2 → 76.8)
+5.9 accuracy on Bn-MSVAMP benchmark (70.5 → 76.4)
88.71% Bengali reasoning (vs 14.79% for base model)
79.5% fewer tokens in generated solutions (943 → 193 words)

Model Overview

Property	Value
Model Type	Causal Language Model
Base Model	Qwen/Qwen3-4B
Parameters	4B
Training	SFT + Curriculum-GRPO
Context Length	4,096 tokens
Language	Bengali, English

Training Details

This model was trained using our multi-stage pipeline:

Supervised Fine-Tuning (SFT): Trained on GANIT-SFT (~11k examples) to ground reasoning in Bengali
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

from transformers import AutoModelForCausalLM, AutoTokenizer

model_name = "dipta007/GanitLLM-4B_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

vllm serve dipta007/GanitLLM-4B_SFT_CGRPO --max-model-len 4096

Performance

Model	Bn-MGSM	Bn-MSVAMP	Avg. Words	Bengali %
Qwen3-4B (base)	69.20	70.50	943	14.79%
GanitLLM-4B_SFT_CGRPO	76.80	76.40	193	88.71%

Related Models

Model	Parameters	Training	Link
GanitLLM-4B_SFT_CGRPO	4B	SFT + CGRPO	Link
GanitLLM-4B_SFT_GRPO	4B	SFT + GRPO	Link
GanitLLM-4B_CGRPO	4B	CGRPO	Link
GanitLLM-1.7B_SFT_CGRPO	1.7B	SFT + CGRPO	Link
GanitLLM-0.6B_SFT_CGRPO	0.6B	SFT + CGRPO	Link

Citation

@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.