| # Streaming RL Persistent Worker 3 | |
| - Updated: `2026-07-25T10:35:40.108303+00:00` | |
| - Queue: `diffusion-training-acceleration` | |
| - Lane: `causal` | |
| - Current experiment: `srl_w03_defassign200` | |
| - Current status: `failed` | |
| | Experiment | Status | Return code | Latest log step | | |
| |---|---|---:|---:| | |
| | `srl_w03_zero10` | completed | 0 | | | |
| | `srl_w03_shuffle20` | completed | 0 | | | |
| | `srl_w03_assign100` | failed | 1 | | | |
| | `srl_w03_zero100` | failed | 1 | | | |
| | `srl_w03_shuffle100` | failed | 1 | | | |
| | `srl_w03_assign200` | failed | 1 | | | |
| | `srl_w03_tail100` | failed | 1 | | | |
| | `srl_w03_defassign200` | failed | 1 | | | |
| Rollout manifests and compact evaluations are mirrored to Hugging Face. | |
| Trajectory tensors remain node-local; checkpoints and timers remain on S3. | |