fix: use exec in endpointry.sh to handle signal correctly via uvicorn #94
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
XciD
approved these changes
Oct 23, 2024
alvarobartt
approved these changes
Oct 28, 2024
alvarobartt
added a commit
to huggingface/Google-Cloud-Containers
that referenced
this pull request
Oct 28, 2024
Kudos to @co42 for the catch at huggingface/huggingface-inference-toolkit#94
alvarobartt
added a commit
to huggingface/Google-Cloud-Containers
that referenced
this pull request
Oct 30, 2024
* Add `pytorch/inference/gpu/2.3.1/transformers/4.46.0/py311` (WIP) - Include missing `requirements.txt` installation in `entrypoint.sh` (required to install custom dependencies with custom models) - Fix Python 3.11 installation as it was not properly installed and Python 3.10 was used instead - Use `uv` to install the dependencies as it's way faster than default `pip` - Also `uv` is able to successfully install `kenlm` which is a `transformers` dependency that `pip` is not able to install when building the `Dockerfile` - Tested with some of the latest models that those bumped dependencies support as Gemma2, Llama3.2, StableDiffusion 3.5, and much more * Remove `uv` and don't upgrade `setuptools` Just by fixing the Python 3.11 and the `pip` installation, the installation issue affecting `kenlm` is solved already; so no need to add `uv` for the moment even though it would be a nice addition * Add `pytorch/inference/cpu/2.3.1/transformers/4.46.0/py311` * Update `pip install` syntax when installing from URL * Add `exec` to `uvicorn` in `entrypoint.sh` Kudos to @co42 for the catch at huggingface/huggingface-inference-toolkit#94 * Remove extra line-break in `Dockerfile` * Update `HF_INFERENCE_TOOLKIT_VERSION` to 0.5.1 See the latest `huggingface-inference-toolkit` release at https://github.com/huggingface/huggingface-inference-toolkit/releases/tag/0.5.1 * Bump `transformers` to 4.46.1 in `huggingface-inference-toolkit` `transformers` 4.46.0 was yanked because Python 3.8 support was unintentionally dropped, whilst also fixes some issues affecting both `torch.fx` and `onnx` Co-authored-by: Philipp Schmid <[email protected]> --------- Co-authored-by: Philipp Schmid <[email protected]>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Right now when a pod running this container is deleted a SIGTERM is sent to
entrypoint.shwhich is not forwarded to uvicorn.There is no graceful shutdown and it will continue to run until a SIGKILL stop it abruptly.