vh 750675e391 feat(intern-decision): cap 9.0 GiB with MAX_TOKENS 7168, the largest call measured to fit ...

Both are required in compose because they are coupled: MAX_TOKENS is checked before the forward
pass, so an oversized call is a clear 422 instead of reaching the cap as a 503. Pre-deploy floor
is nvidia-smi Free >= 15,400 MiB on GPU 1 (card peak 9,876 + scriberr 5,496).

2026-09-30 09:40:08 -07:00

..

acceptance

feat(intern-decision): cap 9.0 GiB with MAX_TOKENS 7168, the largest call measured to fit

2026-09-30 09:40:08 -07:00

src/intern_decision_serve

fix(intern-decision-serve): code-review fixes

2026-09-30 09:30:40 -07:00

tests

fix(intern-decision-serve): code-review fixes

2026-09-30 09:30:40 -07:00

.dockerignore

feat(intern-decision-serve): Intern-Decision-4B behind semif-serve's HTTP surface

2026-09-30 09:04:39 -07:00

.gitignore

feat(intern-decision-serve): Intern-Decision-4B behind semif-serve's HTTP surface

2026-09-30 09:04:39 -07:00

Dockerfile

feat(intern-decision-serve): Intern-Decision-4B behind semif-serve's HTTP surface

2026-09-30 09:04:39 -07:00

intern-decision-serve.contract.md

fix(intern-decision-serve): code-review fixes

2026-09-30 09:30:40 -07:00

pyproject.toml

feat(intern-decision-serve): Intern-Decision-4B behind semif-serve's HTTP surface

2026-09-30 09:04:39 -07:00

uv.lock

feat(intern-decision-serve): Intern-Decision-4B behind semif-serve's HTTP surface

2026-09-30 09:04:39 -07:00