Modelpublic

SeongryongJung/Qwen3-4B-Physics-RLSD-TR

sourceHugging Faceapache-2.0updated 3mo agoView on Hugging Face

0likes9downloads

Model Card

Qwen3-4B-Physics-RLSD-TR

This repository contains the Qwen3-4B Physics RLSD_TR batch-size-32 run.

The repository root contains the best validation checkpoint, selected by validation mean@16. checkpoints/last/ contains the final checkpoint. For this run, best and final are both global_step_100.

Performance

Dataset	Method	Base model	Train batch size	Best val mean@16	Best checkpoint	Final val mean@16	Final checkpoint
Physics	RLSD_TR	Qwen3-4B	32	71.09%	100	71.09%	100

[image]

Validation Mean@16

step	val_mean16	percent
10	0.596875000000	59.69%
20	0.600000000000	60.00%
30	0.624218750000	62.42%
40	0.625000000000	62.50%
50	0.639843750000	63.98%
60	0.622656250000	62.27%
70	0.643750000000	64.38%
80	0.673437500000	67.34%
90	0.692187500000	69.22%
100	0.710937500000	71.09%

Detailed Training Hyperparameters

Section	Parameter	Value	Source
Run identity	`Base model`	`Qwen/Qwen3-4B`	queue/script override
Run identity	`Dataset`	`Physics / SciKnowEval physics`	runqwen3generalization.sh
Run identity	`Method`	`RLSD_TR`	runqwen3generalization.sh
Run identity	`Config`	`rlsd`	runqwen3generalization.sh
Run identity	`Experiment`	`qwen3gen-physics-RLSD_TR-Qwen-Qwen3-4B-mbs8-decay0-tr0.1-train32-rollout8-lr1e-6-vllm0.8`	runqwen3generalization.sh
Run identity	`W&B run`	`run-20260702_085822-eajdt9hv`	wandb
Data	`Train file`	`datasets/sciknoweval/physics/train.parquet`	script override
Data	`Validation file`	`datasets/sciknoweval/physics/test.parquet`	script override
Data	`Train batch size`	`32`	queue/script override
Data	`Train max samples`	`3200`	queue/script override
Data	`Prompt key`	`prompt`	legacy_data.yaml default
Data	`Reward key`	`data_source`	legacy_data.yaml default
Data	`Shuffle train data`	`True`	user.yaml / legacy_data.yaml
Data	`Validation shuffle`	`False`	legacy_data.yaml default
Data	`Filter overlong prompts`	`True`	user.yaml
Data	`Prompt truncation`	`error`	legacy_data.yaml default
Data	`enable_thinking`	`false`	script override
Schedule	`Total training steps`	`100`	queue/script override
Schedule	`Total epochs`	`30`	ppo_trainer/user.yaml default
Schedule	`Validation before train`	`False`	queue/script override
Schedule	`Save frequency`	`10`	queue/script override
Schedule	`Validation frequency`	`10`	queue/script override
Sequence	`Max prompt length`	`2048`	queue/script override
Sequence	`Max response length`	`8192`	queue/script override
Sequence	`Max model length`	`10240`	queue/script override
Sequence	`Actor max token length per GPU`	`10240`	queue/script override
Rollout	`Rollout engine`	`vllm`	user.yaml
Rollout	`Rollout dtype`	`bfloat16`	rollout.yaml default
Rollout	`Train rollout n`	`8`	queue/script override
Rollout	`Train rollout temperature`	`1.0`	script override
Rollout	`Train rollout top_p`	`1.0`	script override
Rollout	`Train rollout do_sample`	`True`	rollout.yaml default
Rollout	`Calculate rollout log probs`	`True`	rlsd.yaml / script override
Rollout	`Max num batched tokens`	`10240`	queue/script override
Rollout	`vLLM GPU memory utilization`	`0.8`	queue/script override
Rollout	`Tensor model parallel size`	`2`	rollout.yaml default
Rollout	`Free cache engine`	`True`	rollout.yaml default
Validation	`Validation rollout n`	`16`	queue/script override
Validation	`Validation temperature`	`0.6`	queue/script override
Validation	`Validation top_p`	`0.95`	queue/script override
Validation	`Validation do_sample`	`True`	queue/script override
Optimization	`Optimizer`	`AdamW`	fsdp optimizer config
Optimization	`Learning rate`	`1e-6`	RLSD_TR method override
Optimization	`LR scheduler`	`constant`	W&B config
Optimization	`LR warmup steps`	`10`	script override
Optimization	`Weight decay`	`0.01`	script override
Optimization	`Betas`	`(0.9, 0.999)`	W&B config
Optimization	`Gradient clip`	`1.0`	script override
PPO/GRPO	`Advantage estimator`	`grpo`	rlsd.yaml
PPO/GRPO	`Normalize GRPO advantages by std`	`False`	script override
PPO/GRPO	`PPO epochs`	`1`	W&B config
PPO/GRPO	`PPO mini batch size`	`8`	queue/script override
PPO/GRPO	`PPO micro batch size per GPU`	`1`	user.yaml
PPO/GRPO	`Clip ratio low`	`0.2`	script override
PPO/GRPO	`Clip ratio high`	`0.28`	script override
PPO/GRPO	`Gamma`	`1.0`	ppo_trainer.yaml default
PPO/GRPO	`Lambda`	`1.0`	ppo_trainer.yaml default
PPO/GRPO	`Use KL in reward`	`False`	ppo_trainer/user.yaml
PPO/GRPO	`Actor KL loss observed`	`0.0`	output.log
Rollout correction	`Importance sampling mode`	`token`	script override
Rollout correction	`IS threshold`	`2.0`	script override
RLSD_TR	`Policy loss mode`	`rlsd`	method override
RLSD_TR	`Teacher regularization`	`trust-region`	method override
RLSD_TR	`Trust-region mix / teacher update rate`	`0.1`	queue/script override
RLSD_TR	`Token reweight lambda`	`0.5`	queue/script override
RLSD_TR	`Token reweight eps_w`	`0.2`	queue/script override
RLSD_TR	`Token reweight decay steps`	`0`	queue/script override
RLSD_TR	`Max reprompt length`	`10240`	method override
RLSD_TR	`Fused kernels`	`False`	method override
FSDP/System	`Actor strategy`	`fsdp`	dp_actor.yaml
FSDP/System	`FSDP dtype`	`bfloat16`	W&B config
FSDP/System	`FSDP model dtype`	`fp32`	W&B config
FSDP/System	`Use torch compile`	`True`	W&B config
FSDP/System	`GPUs per node`	`8`	queue/script override
FSDP/System	`Nodes`	`1`	user.yaml
FSDP/System	`GPU type`	`NVIDIA H200`	wandb-metadata
Checkpoint/Logging	`Checkpoint root`	`checkpoints/datasets/sciknoweval/physics`	script override
Checkpoint/Logging	`Latest checkpointed iteration`	`100`	latestcheckpointediteration.txt
Checkpoint/Logging	`Max actor checkpoints to keep`	`1`	user.yaml
Checkpoint/Logging	`Logger`	`console, wandb`	ppo_trainer.yaml
Checkpoint/Logging	`W&B entity`	`seongryongjung-chung-ang-university`	environment
Checkpoint/Logging	`W&B project`	`SDPO-root`	user.yaml project_name
Checkpoint/Logging	`W&B group`	`QWEN3-RLSD-TR-GRPO-matched-generalization`	method override

Raw result and artifact files:

—results/validation_mean16.csv
—results/training_scores.csv
—results/hyperparameters.csv
—results/training_score.png
—results/training_score.svg
—artifacts/config.yaml
—artifacts/wandb-summary.json
—artifacts/wandb-metadata.json
—artifacts/output.log
—artifacts/queue.log

Usage

python

from transformers import AutoModelForCausalLM, AutoTokenizer

repo_id = "SeongryongJung/Qwen3-4B-Physics-RLSD-TR"
tokenizer = AutoTokenizer.from_pretrained(repo_id, trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained(
    repo_id,
    torch_dtype="auto",
    device_map="auto",
    trust_remote_code=True,
)

Source

—Checkpoint: checkpoints/datasets/sciknoweval/physics/qwen3gen-physics-RLSD_TR-Qwen-Qwen3-4B-mbs8-decay0-tr0.1-train32-rollout8-lr1e-6-vllm0.8
—W&B run: run-20260702_085822-eajdt9hv
—Queue log: artifacts/queue.log