AstrAI

819 Commits 1 Branch 14 Tags

3 Commits

Author	SHA1	Message	Date
0z5a	90de5bc1bd	fix: report failed rollout requests - Return structured finish and error reasons for synchronous generation - Reject failed online rollout batches instead of training on empty responses - Verify allocation and extension failures release metrics and KV state	2026-09-02 12:52:50 +08:00
0z5a	c36846c8a4	fix: condition online DPO on rollout prompts - Concatenate rollout prompts with selected chosen and rejected responses - Mask prompt tokens from DPO loss while preserving explicit attention visibility - Cover response selection, padding alignment, and prompt-conditioned inputs	2026-09-02 12:52:50 +08:00
0z5a	08721f6d31	fix: save checkpoints after optimizer steps - add a post-step callback hook for checkpoint saves - preserve updated model, optimizer, and scheduler state - cover checkpoint ordering with a regression test	2026-09-01 12:25:27 +08:00

Author

SHA1

Message

Date

0z5a

90de5bc1bd

fix: report failed rollout requests

- Return structured finish and error reasons for synchronous generation
- Reject failed online rollout batches instead of training on empty responses
- Verify allocation and extension failures release metrics and KV state

2026-09-02 12:52:50 +08:00

0z5a

c36846c8a4

fix: condition online DPO on rollout prompts

- Concatenate rollout prompts with selected chosen and rejected responses
- Mask prompt tokens from DPO loss while preserving explicit attention visibility
- Cover response selection, padding alignment, and prompt-conditioned inputs

2026-09-02 12:52:50 +08:00

0z5a

08721f6d31

fix: save checkpoints after optimizer steps

- add a post-step callback hook for checkpoint saves
- preserve updated model, optimizer, and scheduler state
- cover checkpoint ordering with a regression test

2026-09-01 12:25:27 +08:00