# Streaming RL Persistent Worker 9

- Updated: `2026-07-26T01:12:18.903675+00:00`
- Queue: `diffusion-training-acceleration`
- Lane: `boundary`
- Current experiment: `none`
- Current status: `queue_completed`

| Experiment | Status | Return code | Latest log step |
|---|---|---:|---:|
| `srl_w09_balanced20` | completed | 0 |  |
| `srl_w09_deficit20` | completed | 0 |  |
| `srl_w09_deficit100` | failed | 1 |  |
| `srl_w09_random100` | failed | 1 |  |
| `srl_w09_tail100` | failed | 1 |  |
| `srl_w09_last100` | failed | 1 |  |
| `srl_w09_balanced100` | failed | 1 |  |
| `srl_w09_deficit200` | failed | 1 |  |

Rollout manifests and compact evaluations are mirrored to Hugging Face.
Trajectory tensors remain node-local; checkpoints and timers remain on S3.