[fix] Pre-parse data + decouple args

[arekay-nv]

Fixes the dataloader to pre-parse and decouples the args from benchmark execution.

• The existing dataloader was parsing the data on-demand which can slow down the hot loop when sending requests. Instead, we want to pre-parse the data and have them sent quickly.
• Decouples the command line arguments from the benchmark execution. This makes it easier to reproduce executions from a config file since the command line arguments populate a structure which can be directly serialized/deserialized to yaml.

Type of change

• Bug fix
• New feature
• Documentation update
• Refactor/cleanup

Related issues

Testing

• Tests added/updated
• All tests pass locally
• Manual testing completed

Checklist

• Code follows project style
• Pre-commit hooks pass
• Documentation updated (if needed)

[github-actions]

MLCommons CLA bot All contributors have signed the MLCommons CLA ✍️ ✅

[Copilot]

Pull request overview

This PR fixes the dataloader to pre-parse dataset samples and decouples command-line arguments from benchmark execution logic. The pre-parsing optimization moves data parsing out of the hot loop to improve request throughput, while the argument decoupling enables easier config serialization/deserialization for reproducible benchmark runs.

Key Changes:

• Pre-parse all dataset samples during load instead of on-demand parsing
• Store command-line arguments (report_dir, timeout, output, verbose) directly in BenchmarkConfig
• Remove max_concurrency setting from client configuration across all templates and code

Reviewed changes

Copilot reviewed 17 out of 17 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
src/inference_endpoint/dataset_manager/dataloader.py	Implements pre-parsing by iterating through data during load() and storing parsed results
src/inference_endpoint/commands/benchmark.py	Refactors to use config fields instead of args namespace, removes args parameter from _run_benchmark
src/inference_endpoint/config/schema.py	Adds output, report_dir, timeout, verbose fields to BenchmarkConfig; removes max_concurrency validation
src/inference_endpoint/load_generator/session.py	Renames report_path parameter to report_dir for consistency
src/inference_endpoint/cli.py	Renames --report-path CLI argument to --report-dir
tests/unit/commands/test_benchmark.py	Removes max_concurrency assertions and adds new config fields to mock args
tests/unit/config/test_yaml_loader.py	Skips concurrency validation tests pending refactor
tests/integration/*.py	Updates parser functions to access dictionary keys instead of object attributes
tests/conftest.py	Simplifies parser to return row directly instead of reconstructing dictionary
Config templates (*.yaml)	Removes max_concurrency setting from all template files
Example configs (*.yaml)	Removes max_concurrency setting from example configuration files

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

Pull request overview

Copilot reviewed 17 out of 17 changed files in this pull request and generated 2 comments.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.