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.