Samuel Knight

CUDA version in documentation for base 1 is out of date. Images (e.g. [cml:0-dvc2-base1-gpu](https://hub.docker.com/layers/cml/iterativeai/cml/0-dvc2-base1-gpu/images/sha256-a6bc7bc5e4a9c3806682275edcee6235d4ec05b6219bba405e74950f122f8a27?context=explore)) currently use CUDA: 11.2.1 and CUDNN: 8.1.1

Mirrored strategy model.load_weights() failure

8

### Issue type Bug ### Have you reproduced the bug with TensorFlow Nightly? No -- appears to be fixed in Nightly ### Source binary ### TensorFlow version tf 2.15 ###...

When legacy gitlab runner registration token is disabled old flow returns an undefined token which causes the null check to enter the new token code path to fail. Tested runner-token...

The current code for acquiring a runnerToken attempts to fetch a token in the following order: 1. Use the legacy endpoint (`/projects/` 2. Use the new runner authentication endpoint (`/user/runners`)...