---
base_model: laion/GLM-4_7-swesmith-sandboxes-with_tests-oracle_verified_120s-maxeps-131k-fixthink
tags:
- rl
- skyrl
- agentic
- swe
library_name: transformers
---

# ablation-pymethods2test-shaped-45-8B

RL (SkyRL GRPO) checkpoint from the **shaped-reward ablation** of the a3-successor
study. Reward = shaped pass-ratio (fraction of tests passing, `reward_shaper=pass_ratio`),
as opposed to the binary all-tests-pass reward of the a3 series.

- **Base model:** [laion/GLM-4_7-swesmith-sandboxes-with_tests-oracle_verified_120s-maxeps-131k-fixthink](https://huggingface.co/laion/GLM-4_7-swesmith-sandboxes-with_tests-oracle_verified_120s-maxeps-131k-fixthink) (a Qwen3-8B SFT)
- **Training dataset:** [DCAgent/exp_rpt_pymethods2test-large](https://huggingface.co/datasets/DCAgent/exp_rpt_pymethods2test-large)
- **Checkpoint:** `global_step_45`, selected as the best checkpoint by **EMA
  (alpha=1/3, trailing-5 window) of `reward/avg_raw_reward`** computed across
  the full 80-step training chain (EMA = 0.4712 at step 45).
- **Training:** 80 steps total, `hf_save_interval=5`, 14x GH200 nodes on JSC Jupiter.

The `rl_config.json` in this repo is the exact launch config used for reproducibility.

## Training Traces

Training-time Daytona/Harbor rollouts for this run are uploaded as
a companion dataset:
**[penfever/ablation-pymethods2test-shaped](https://huggingface.co/datasets/penfever/ablation-pymethods2test-shaped)**

The dataset contains the `last` episode of each trial (per
`make_and_upload_trace_dataset --episodes last`) — the same rollouts
the policy was trained on after rollback / truncation.