Fix: Eagerly initialize ML Diagnostics to prevent Protobuf descriptor collision crash

[bvandermoon]

Description

Eagerly initializes libtpu_metric at the beginning of MaxText's root initialization to prevent a fatal C++ Protobuf descriptor registration crash.

Context & Problem

Customers experience a fatal C++ Protobuf descriptor registration crash (File already exists in database: google/protobuf/timestamp.proto) when running MaxText training jobs with ML Diagnostics enabled (managed_mldiagnostics=True).

Root Cause

Lazy loading of ML Diagnostics occurs midway through the training flow. By then, JAX and other extensions have already registered timestamp.proto in the shared DescriptorPool. When libtpu_metric._initialize() is called later, it tries to register the same descriptor, causing an abort.

Solution

Eagerly initialize libtpu_metric at the beginning of MaxText's root initialization to ensure registration happens before JAX imports and calls cloud_tpu_init(). We use max_logging for cleaner logging integration and catch any Exception (e.g. if the package isn't installed) to fail gracefully.

Tests

Verified on dedicated TPU VM:

1. Checked that importing maxtext package completes successfully and prints INFO:absl:ML diagnostics initialization without any Protobuf descriptor registration crash:

   python3 -c "from absl import logging; logging.set_verbosity(logging.INFO); import maxtext"

2. Ran all unit tests in tests/unit/configs_test.py and they all passed:

   pytest ../tests/unit/configs_test.py

Checklist

Before submitting this PR, please make sure (put X in square brackets):

• I have performed a self-review of my code. For an optional AI review, add the gemini-review label.
• I have necessary comments in my code, particularly in hard-to-understand areas.
• I have run end-to-end tests tests and provided workload links above if applicable.
• I have made or will make corresponding changes to the doc if needed, including adding new documentation pages to the relevant Table of Contents (toctree directive) as explained in our documentation.

[codecov]

Codecov Report

❌ Patch coverage is 75.00000% with 2 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
src/maxtext/__init__.py	75.00%	2 Missing ⚠️

📢 Thoughts on this report? Let us know!