Skip to content

chore: pin baked hf CLI to 1.16.4 to match framework (no PyPI fetch at boot) #23

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
Johan-de-R merged 1 commit into from
Jul 14, 2026
Merged

chore: pin baked hf CLI to 1.16.4 to match framework (no PyPI fetch at boot) #23

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
20 changes: 11 additions & 9 deletions examples/cosmos3/Dockerfile
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -63,10 +63,12 @@ ENV TRITON_PTXAS_PATH="/usr/local/cuda/bin/ptxas"
RUN --mount=type=cache,target=/root/.cache/uv \
uv pip install fastapi uvicorn ai-dynamo==1.1.1 uvloop

# Pre-install the HF CLI tool so that `uvx hf@1.13.0` reuses this
# installation instead of creating an ephemeral env missing `click`.
# Pre-install the HF CLI tool at the version the framework invokes at runtime
# (`uvx hf@1.16.4` in cosmos_framework's checkpoint_db) so uvx reuses this
# baked copy instead of fetching the tool from PyPI at every pod boot. MUST
# match the framework's pinned hf version, else uvx re-downloads it.
RUN --mount=type=cache,target=/root/.cache/uv \
uv tool install hf@1.13.0 --with click
uv tool install hf@1.16.4 --with click

RUN --mount=type=cache,target=/root/.cache/uv \
uv pip install --no-deps -e /workspace
Expand All @@ -83,32 +85,32 @@ RUN --mount=type=cache,target=/root/.cache/uv \
# — guardrails are disabled in worker.py (guardrails=False), so it is never
# loaded at runtime. Re-add its download here (and the BuildKit hf_token secret
# it needs) if guardrails are ever turned back on.
RUN uvx hf@1.13.0 download --format=json \
RUN uvx hf@1.16.4 download --format=json \
Qwen/Qwen3-0.6B \
--repo-type model \
--revision c1899de289a04d12100db370d81485cdf75e47ca \
--include '*.json' --include '*.txt' && \
uvx hf@1.13.0 download --format=json \
uvx hf@1.16.4 download --format=json \
Qwen/Qwen3-VL-2B-Instruct \
--repo-type model \
--revision 89644892e4d85e24eaac8bacfd4f463576704203 \
--include '*.json' --include '*.txt' && \
uvx hf@1.13.0 download --format=json \
uvx hf@1.16.4 download --format=json \
Qwen/Qwen3-VL-8B-Instruct \
--repo-type model \
--revision 0c351dd01ed87e9c1b53cbc748cba10e6187ff3b \
--include '*.json' --include '*.txt' && \
uvx hf@1.13.0 download --format=json \
uvx hf@1.16.4 download --format=json \
Qwen/Qwen3-VL-32B-Instruct \
--repo-type model \
--revision 0cfaf48183f594c314753d30a4c4974bc75f3ccb \
--include '*.json' --include '*.txt' && \
uvx hf@1.13.0 download --format=json \
uvx hf@1.16.4 download --format=json \
Wan-AI/Wan2.1-T2V-14B \
--repo-type model \
--revision a064a6c71f5be440641209c07bf2a5ce7a2ff5e4 \
Wan2.1_VAE.pth && \
uvx hf@1.13.0 download --format=json \
uvx hf@1.16.4 download --format=json \
Wan-AI/Wan2.2-TI2V-5B \
--repo-type model \
--revision 921dbaf3f1674a56f47e83fb80a34bac8a8f203e \
Expand Down
Loading