Skip to content

5080deploy: merge upstream #108, GET /v1/models/{model_id} - #61

Open
gdevenyi wants to merge 2 commits into
5080/22-prefill-invalidation-500from
5080/23-retrieve-model-108
Open

gdevenyi wants to merge 2 commits into
5080/22-prefill-invalidation-500from
5080/23-retrieve-model-108

Conversation

@gdevenyi

Copy link
Copy Markdown
Owner

Stacked on #60. Merges upstream FlashML-org#108 (head 637aa87): OpenAI's retrieve-model operation, GET /v1/models/{model_id}, which clients such as Visual Studio call after listing models and treat a 404 as "model not loaded".

Merge: clean.

Testing (RTX 5080, production flags):

  • Live: GET /v1/models/Qwen3.8-Flash-Next-NVFP4 -> 200 with context_length 262144; an unknown id -> 404.
  • pytest tests/ -m "not slow" at the top of this stack (77d8d5c): 2446 passed, 206 skipped; the 4 failures are the known environment-only ones (3x test_muse_glimmer_vision.py on transformers 5.15.1, test_async_pull_coalesced needing pytest-asyncio).

🤖 Generated with Claude Code

https://claude.ai/code/session_01Bu6LgoxLR4wETqb7RPR2vt

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.

2 participants