Qwen3-4B-GRPO-math-reasoning

This model is a fine-tuned version of Qwen3-4B using GRPO (Group Relative Policy Optimization) without KL penalty for mathematical reasoning.

Trained with PipelineRL.

Training Details

Datasets

Split	Datasets
Train	gsm8k_train, math_train
Test	gsm8k_test, math_500

RL Algorithm

Parameter	Value
Algorithm	GRPO (Group Relative Policy Optimization)
Policy Loss	ppo
KL Coefficient	0.0
Epsilon (clip)	0.02
Divide Advantage by Std	False
Filter Zero Advantage Groups	False
Rollouts per Problem	16

Training Hyperparameters

Parameter	Value
Base Model	Qwen/Qwen3-4B
Learning Rate	1e-06
LR Scheduler	cosine
Warmup Steps	25
Max Training Steps	1500
Micro Batch Size	2
Gradient Accumulation	128
Effective Batch Size	256
Sequence Length	8192
Gradient Clipping	0.3
Weight Decay	0.01
Optimizer	adamw_torch
Precision	bf16
DeepSpeed	ZeRO Stage 3

Evaluation Results

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

Dataset	pass@1	pass@2	pass@4	pass@8	pass@16	pass@32
GSM8K (test)	89.11	91.69	93.33	94.36	95.03	95.53
MATH-500	79.90	85.70	89.76	92.62	94.65	96.00
Overall	86.58	90.05	92.35	93.88	94.93	95.66

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_grpo_no_kl_3a1f_4xh100_197341_finetune_33839739

Usage

Transformers

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("jaygala24/Qwen3-4B-GRPO-math-reasoning", revision="step-0200")  # or whatever branch name, e.g. "step-0400", "step-0600"
tokenizer = AutoTokenizer.from_pretrained("jaygala24/Qwen3-4B-GRPO-math-reasoning", revision="step-0200")  # or whatever branch name, e.g. "step-0400", "step-0600"

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-GRPO-math-reasoning", revision="step-0200")  # or whatever branch name, e.g. "step-0400", "step-0600"
sampling_params = SamplingParams(temperature=0.7, max_tokens=4096)

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