Add ray to each benchmarking entry and make sure no benchmarking/script has RayClient

[praateekmahajan]

1. Added missing ray: config (num_cpus: 64, num_gpus: 4) to:
   • domain_classification_xenna
   • embedding_generation_xenna
   • audio_fleurs
2. Removed RayClient import, ray_client.start(), and ray_client.stop(). The Ray cluster is already started by run.py.
   • image_pipeline_benchmark.py
   • multimodal_mint1t_benchmark.py
3. Fixed success_code = 0 (should be 1) in arxiv_e2e_pipeline_benchmark.py which masked benchmark failures as successes
4. Replaced hardcoded XennaExecutor() in audio_fleurs_benchmark.py with setup_executor() + --executor CLI arg, consistent with all other benchmark scripts
5. Fixed potential UnboundLocalError crashes across multiple scripts (exact_dedup_identification, score_filter, modifier, video_pipeline, image_pipeline, fasttext_filter, semdedup_identification, dedup_removal) by initializing variables before conditional/try blocks

Description

Usage

# Add snippet demonstrating usage

Checklist

• I am familiar with the Contributing Guide.
• New or Existing tests cover these changes.
• The documentation is up to date with these changes.

[greptile-apps]

Unhandled NameError can propagate past finally

run_exact_duplicate_identification_benchmark() has an internal except Exception block that sets success = False, but it does not define workflow_result when the exception occurs before line 77 (workflow_result = workflow.run(...)). The return at line 118 then references workflow_result, causing a NameError that escapes the inner function.

Previously, the outer except Exception in main() would have caught this and written clean fallback results. With the new try/finally-only pattern, the NameError propagates past finally (fallback results are still written correctly), but main() never returns — it exits via an unhandled Python exception traceback instead of a clean exit code.

If an unambiguous exit code is desired even in this edge case, consider adding a narrow guard in the inner function:

return {
    ...
    "tasks": workflow_result if success else [],
}

[greptile-apps]

num_gpus: 4 vs actual GPU usage — worth verifying

domain_classification_xenna and embedding_generation_xenna (line 220-223) now both reserve 4 GPUs on the Ray cluster. Neither entry passes a --gpus or similar argument to their scripts, so it isn't immediately obvious how many GPUs the workloads actually consume at runtime.

If these benchmarks only use 1 GPU internally (similar to the audio_fleurs case), 3 GPUs per run are reserved but idle for the entire benchmark duration, potentially blocking other concurrent jobs. It may be worth confirming the per-script GPU allocation before landing this so the reservation matches actual utilisation.

This same concern applies to the embedding_generation_xenna entry at line 220.

[praateekmahajan]

Currently it anyway uses 4 gpus we are just now being explicit it.

[ayushdg]

Thanks!

[ayushdg]

One nice thing about throughput requirements is that we don't need to update with numbers GPUs.

[ayushdg]

unrelated to the PR but this arg is confusing IMO. The Cli claims its the number of GPUs to use but actually its the resource requirement for the inference stage