Colab-ready GPU validation for ArtP and DArtP

[jsalsman]

Summary

This PR adds a reproducible GPU setup and smoke-test workflow for PathBench’s ArtP and DArtP evaluators, including support for Google Colab T4 runtimes.

The new tools/test_gpu_predictors.py helper:

• creates and reuses an isolated Python virtual environment;
• installs pinned CUDA-enabled PyTorch and torchaudio packages;
• verifies that PyTorch can access the NVIDIA GPU;
• installs PathBench and its test dependencies;
• runs focused ArtP and DArtP smoke tests;
• uses a headless Matplotlib backend in notebook environments;
• optionally retrieves the English KenLM model from the existing Zenodo archive using HTTP byte ranges, avoiding a download of the complete 35 GB archive;
• validates the downloaded model using its compressed and expanded sizes, ZIP CRC-32, SHA-256, and KenLM;
• reuses an existing verified model on subsequent runs;
• retains the previous skip behavior when the optional language-model download is not requested.

The dependency configuration now uses the published phonemizer-fork==3.3.2 release instead of an inaccessible Git repository.

Documentation and network-free tests cover the GPU workflow, range-based ZIP extraction, archive validation, checksum failures, interrupted requests, insufficient disk space, cache reuse, and related failure cases. An opt-in live test verifies that the expected English model remains available at the pinned location in the Zenodo archive.

Colab validation

Proof: https://colab.research.google.com/drive/1o3G8XjImfuzn2mpvtDrl3_7Y4LEixsfL

The cumulative changes were tested in Google Colab with:

• Python 3.12.14
• T4 GPU
• NVIDIA driver 580.82.07
• PyTorch 2.6.0+cu124
• PyTorch CUDA runtime 12.4
• wiki_en_token.arpa.bin size: 14,600,342,241 bytes

Results:

• ArtP smoke test: passed
• DArtP smoke test: passed
• GPU-helper unit suite: 33 passed, 1 skipped

The skipped default unit test is the intentionally opt-in live-network check.

This upstream PR consolidates the work developed and validated through PRs #1–#6 in the fork.

[jsalsman]

Premature; forgot to include:

• install Ubuntu build packages;
• build and install the pinned espeak-ng;
• install or configure the NVIDIA driver