Skip to content

fix(utils): read stop ids from config.json without generation_config - #874

Merged
jason-fxz merged 1 commit into
mainfrom
eos-config-fallback
Nov 6, 2026
Merged

jason-fxz merged 1 commit into
mainfrom
eos-config-fallback

Conversation

@jason-fxz

Copy link
Copy Markdown
Collaborator

load_eos_token_ids now reads eos_token_id from config.json when generation_config.json is missing, so ricursiveAI/Ripple-65B-Preview (#871) also stops on <|endoftext|>. Checkpoints with the file keep their stop set.

@jason-fxz
jason-fxz merged commit e5856ac into main Nov 6, 2026
1 check passed
pretor pushed a commit to pretor/FreeToken that referenced this pull request Nov 6, 2026
KarrAcaRn pushed a commit to KarrAcaRn/FreeToken-ByAI that referenced this pull request Nov 7, 2026
Upstream 14119b8 (FlashML-org#860 FlashML-org#867 FlashML-org#868 FlashML-org#872 FlashML-org#873 FlashML-org#866 FlashML-org#874 FlashML-org#871).

Conflicts:
- server/generation.py, server/openai_api.py: upstream FlashML-org#868 moves the
  stream token counts into GenUsage (continuous_usage_stats). Kept next's
  GenTimer/gen_timings, chat logprobs and the n>1 completion fan-out (FlashML-org#843)
  and moved their counters onto GenUsage, so continuous usage also covers
  every choice of a fanned-out completion stream; the metrics block reads
  the same GenUsage.
- tests/server/test_openai_api.py: FlashML-org#868's completions case called
  stream_completion_chunks(uid, ...); on next it goes through
  handle_completion, which builds the fan-out.
- docs/models.md: upstream's Ripple row and GGUF table, next's Flint link.
- README.md: next's fork overview.

Assisted-by: Claude Opus 5.5
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant