Test/generate unit tests #204
Conversation
- Add return type hint to get_tokenizer() - Add type hints and checkpoint validation to generate.py main() - Add parameter type hints to suppress_output() in torch/utils.py Improves IDE support and catches potential bugs early.
- Test argument parsing with all options - Test torch/triton/vllm backend initialization - Test checkpoint path validation - Test generation flow with/without limits - Test tokenizer integration
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
| def main(args: argparse.Namespace) -> None: | ||
| # Validate checkpoint path exists | ||
| checkpoint_path = Path(args.checkpoint) | ||
| if not checkpoint_path.exists(): | ||
| raise FileNotFoundError(f"Checkpoint path does not exist: {args.checkpoint}") |
There was a problem hiding this comment.
The new guard raises FileNotFoundError whenever args.checkpoint is not an existing local path before any backend-specific logic runs. The vLLM backend previously accepted Hugging Face model IDs or other remote identifiers and let VLLMGenerator download the weights itself. With this change, calling python -m gpt_oss.generate meta-llama/Meta-Llama-3-8B-Instruct -b vllm now fails immediately even though the backend could handle it. Consider limiting the existence check to backends that truly require local files (torch/triton) or skipping it for identifiers that are not paths.
Useful? React with 👍 / 👎.
unit tests for generate.py backend selection and argument parsing.
Tests Added: