Upgrade Colab base image #1533
Upgrade Colab base image #1533
Conversation
From `20251024-060052_RC00` to `20260126-060048_RC00`. Added justification to the diff about other changes included. http://b/331651883
| RUN uv pip install --no-build-isolation --no-cache --system "git+https://github.com/Kaggle/learntools" | ||
|
|
||
| # newer daal4py requires tbb>=2022, but libpysal is downgrading it for some reason | ||
| RUN uv pip install --system --no-cache "tbb>=2022" "libpysal==4.9.2" |
There was a problem hiding this comment.
the new colab base image includes tbb 2022.3.0 & libpysal 4.14.1. No reinstall needed.
|
|
||
| # b/404590350: Ray and torchtune have conflicting tune cli, we will prioritize torchtune. | ||
| # b/315753846: Unpin translate package, currently conflicts with adk 1.17.0 | ||
| # b/468379293: Unpin Pandas once cuml/cudf are compatible, version 3.0 causes issues |
There was a problem hiding this comment.
numpy/cudf: The pinned version are the ones now installed in the colab base image.
| # b/468383498: numpy will auto-upgrade to 2.4.x, which causes issues with numerous packages | ||
| # b/404590350: Ray and torchtune have conflicting cli named `tune`. `ray` is not part of Colab's base image. Re-install `tune` to ensure the torchtune CLI is available by default. | ||
| # b/468367647: Unpin protobuf, version greater than v5.29.5 causes issues with numerous packages | ||
| RUN uv pip install --system --force-reinstall --no-cache --no-deps torchtune "google-cloud-translate==3.12.1" "numpy==2.0.2" "pandas==2.2.2" |
There was a problem hiding this comment.
google-cloud-translate 3.24.0 and adk 1.21.0 are now installed in the new Colab base image.
| scikit-surprise | ||
| git+https://github.com/facebookresearch/segment-anything.git | ||
| squarify | ||
| tensorflow-cloud |
There was a problem hiding this comment.
This has been unmaintained since 2021
| tensorflow-cloud | ||
| tensorflow-io | ||
| # Must be compatible with torch version: https://github.com/meta-pytorch/torchcodec?tab=readme-ov-file#installing-torchcodec | ||
| torchcodec==0.9 |
There was a problem hiding this comment.
Starting in 2.9.0, torchaudio.load() uses torchcodec which needs to be installed separately.
See: https://docs.pytorch.org/audio/2.8/generated/torchaudio.load_with_torchcodec.html
| client = UserSecretsClient() | ||
| secret_response = client._get_cloudai_access_token() | ||
| self.assertEqual(secret_response, (secret, now + timedelta(seconds=3600))) | ||
| def call_get_translation_access_token(): |
There was a problem hiding this comment.
These 4 methods were not called (dead code). The matching client._get_*_access_token() don't exist anymore. The AI Platform credentials are now used for accessing these services.
1 participant
From
20251024-060052_RC00to20260126-060048_RC00.Added justification to the diff about other changes included.
Next: Move the CPU image to use the Colab dedicated CPU base image.
http://b/331651883