43 lines
1.2 KiB
Markdown
43 lines
1.2 KiB
Markdown
---
|
|
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-Tooluse-RLSD-TR
|
|
|
|
This repository contains the Qwen3-8B `tooluse` `RLSD_TR` batch-size-32 run.
|
|
|
|
## Performance
|
|
|
|
| Dataset | Method | Base model | Train batch size | Best val mean@16 | Best checkpoint | Final val mean@16 | Final checkpoint |
|
|
|---|---|---|---:|---:|---:|---:|---:|
|
|
| Tool-use / tooluse | RLSD_TR | Qwen3-8B | 32 | 63.42% | 100 | 63.42% | 100 |
|
|
|
|
## Validation Mean@16
|
|
|
|
| step | val_mean16 | percent |
|
|
|---:|---:|---:|---:|
|
|
| 10 | 0.572610294118 | 57.26% |
|
|
| 20 | 0.600183823529 | 60.02% |
|
|
| 30 | 0.588235294118 | 58.82% |
|
|
| 40 | 0.605698529412 | 60.57% |
|
|
| 50 | 0.619485294118 | 61.95% |
|
|
| 60 | 0.614889705882 | 61.49% |
|
|
| 70 | 0.602022058824 | 60.20% |
|
|
| 80 | 0.610294117647 | 61.03% |
|
|
| 90 | 0.625919117647 | 62.59% |
|
|
| 100 | 0.634191176471 | 63.42% |
|
|
|
|
## Source
|
|
|
|
- Checkpoint: `/mnt/mole/SDPO/L2T/checkpoints/datasets/tooluse/qwen3gen-tooluse-RLSD_TR-Qwen-Qwen3-8B-mbs8-decay0-tr0.1-train32-rollout8-lr1e-6-vllm0.8/global_step_100/actor`
|
|
- W&B run: `run-20260703_141322-che709vc`
|