Qwen3-4B-DAPO-math-reasoning

Qwen3-4B-DAPO-math-reasoning

This model is a fine-tuned version of Qwen3-4B using DAPO (Decoupled Clip and Dynamic Sampling Policy Optimization) without KL penalty for mathematical reasoning.

Trained with PipelineRL.

Training Details

Datasets

RL Algorithm

DAPO extends GRPO with clip-higher (asymmetric PPO clipping), dynamic sampling (filtering zero-variance groups), token-level loss aggregation, and overlong reward shaping.

Training Hyperparameters

Evaluation Results

Pass@k on math reasoning benchmarks (N=32 samples per problem, temperature=1.0):

GSM8K test: 1319 problems · MATH-500: 500 problems · Overall: 1819 problems (overall weighted by problem count).

Training Curves

W&B Run

Full training logs: https://wandb.ai/jaygala24-team/rl-post-training/runs/qwen3_4b_dapo_no_kl_3a1f_4xh100_236317_finetune_fd97f02e

Usage

Transformers

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("jaygala24/Qwen3-4B-DAPO-math-reasoning", revision="step-0200")  # optional branch, e.g. "step-0400"
tokenizer = AutoTokenizer.from_pretrained("jaygala24/Qwen3-4B-DAPO-math-reasoning", revision="step-0200")

prompt = "Please reason step by step, and put your final answer within \\boxed{}.\n\nWhat is the sum of 123 and 456?"
inputs = tokenizer(prompt, return_tensors="pt")
outputs = model.generate(**inputs, max_new_tokens=4096, temperature=0.7)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))

vLLM

from vllm import LLM, SamplingParams

llm = LLM(model="jaygala24/Qwen3-4B-DAPO-math-reasoning", revision="step-0200")  # optional branch, e.g. "step-0400"
sampling_params = SamplingParams(temperature=0.7, max_tokens=4096)

prompt = "Please reason step by step, and put your final answer within \boxed{}.

What is the sum of 123 and 456?"
outputs = llm.generate([prompt], sampling_params)
print(outputs[0].outputs[0].text)

Framework

• PipelineRL
• Transformers
• DeepSpeed (ZeRO Stage 3)