Qwen3-8B-Chemistry-RLSD-TR/README.md

---
license: apache-2.0
library_name: transformers
pipeline_tag: text-generation
tags:
- qwen3
- reinforcement-learning
- rlsd-tr
- text-generation
- trust-region
base_model: Qwen/Qwen3-8B
---

# Qwen3-8B-Chemistry-RLSD-TR

This repository contains the Qwen3-8B `chemistry` `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.

## Performance

| Dataset | Method | Base model | Train batch size | Best val mean@16 | Best checkpoint | Final val mean@16 | Final checkpoint |
|---|---|---|---:|---:|---:|---:|---:|
| Chemistry / SciKnowEval chemistry | RLSD_TR | Qwen3-8B | 32 | 70.98% | 80 | 70.65% | 100 |

![Training and validation scores](results/training_score.png)

## Validation Mean@16

| step | val_mean16 | percent |
|---:|---:|---:|
| 10 | 0.421726190476 | 42.17% |
| 20 | 0.519642857143 | 51.96% |
| 30 | 0.606845238095 | 60.68% |
| 40 | 0.645535714286 | 64.55% |
| 50 | 0.668750000000 | 66.88% |
| 60 | 0.687500000000 | 68.75% |
| 70 | 0.696726190476 | 69.67% |
| 80 | 0.709821428571 | 70.98% |
| 90 | 0.709226190476 | 70.92% |
| 100 | 0.706547619048 | 70.65% |

## Detailed Training Hyperparameters

| Section | Parameter | Value | Source |
|---|---|---:|---|
| Run identity | `Base model` | `Qwen/Qwen3-8B` | queue/script override |
| Run identity | `Dataset` | `Chemistry / SciKnowEval chemistry` | run_qwen3_generalization.sh |
| Run identity | `Method` | `RLSD_TR` | run_qwen3_generalization.sh |
| Run identity | `Config` | `rlsd` | run_qwen3_generalization.sh |
| Run identity | `Experiment` | `qwen3gen-chemistry-RLSD_TR-Qwen-Qwen3-8B-mbs8-decay0-tr0.1-train32-rollout8-lr1e-6-vllm0.8` | run_qwen3_generalization.sh |
| Run identity | `W&B run` | `run-20260703_014331-zs3h1j56` | wandb |
| Data | `Train file` | `datasets/sciknoweval/chemistry/train.parquet` | script override |
| Data | `Validation file` | `datasets/sciknoweval/chemistry/test.parquet` | script override |
| Data | `Train batch size` | `32` | queue/script override |
| Data | `Train max samples` | `3200` | queue/script override |
| Schedule | `Total training steps` | `100` | queue/script override |
| 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 |
| Rollout | `Train rollout n` | `8` | queue/script override |
| Rollout | `Validation rollout n` | `16` | queue/script override |
| Rollout | `vLLM GPU memory utilization` | `0.8` | queue/script override |
| Optimization | `Learning rate` | `1e-6` | RLSD_TR method override |
| Optimization | `Weight decay` | `0.01` | script override |
| PPO/GRPO | `PPO mini batch size` | `8` | queue/script override |
| PPO/GRPO | `Normalize GRPO advantages by std` | `False` | baseline_grpo.yaml / script override |
| Rollout correction | `Importance sampling mode` | `token` | script override |
| Rollout correction | `IS threshold` | `2.0` | script override |
| Checkpoint/Logging | `Checkpoint root` | `checkpoints/datasets/sciknoweval/chemistry/qwen3gen-chemistry-RLSD_TR-Qwen-Qwen3-8B-mbs8-decay0-tr0.1-train32-rollout8-lr1e-6-vllm0.8` | script override |
| Checkpoint/Logging | `Latest checkpointed iteration` | `100` | latest_checkpointed_iteration.txt |
| Checkpoint/Logging | `External actor archive` | `checkpoints/datasets/sciknoweval/chemistry/qwen3gen-chemistry-RLSD_TR-Qwen-Qwen3-8B-mbs8-decay0-tr0.1-train32-rollout8-lr1e-6-vllm0.8/_actor_archive` | preserve_actor_checkpoints.py |
| Checkpoint/Logging | `Logger` | `console, wandb` | ppo_trainer.yaml |
| 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 | `Fused kernels` | `False` | 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/output.log`
- `artifacts/queue.log`

## Usage

```python
from transformers import AutoModelForCausalLM, AutoTokenizer

repo_id = "SeongryongJung/Qwen3-8B-Chemistry-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/chemistry/qwen3gen-chemistry-RLSD_TR-Qwen-Qwen3-8B-mbs8-decay0-tr0.1-train32-rollout8-lr1e-6-vllm0.8`
- Root actor checkpoint: `checkpoints/datasets/sciknoweval/chemistry/qwen3gen-chemistry-RLSD_TR-Qwen-Qwen3-8B-mbs8-decay0-tr0.1-train32-rollout8-lr1e-6-vllm0.8/global_step_80/actor`
- Last actor checkpoint: `checkpoints/datasets/sciknoweval/chemistry/qwen3gen-chemistry-RLSD_TR-Qwen-Qwen3-8B-mbs8-decay0-tr0.1-train32-rollout8-lr1e-6-vllm0.8/global_step_100/actor`
- W&B run: `run-20260703_014331-zs3h1j56`
- Queue log: `artifacts/queue.log`