Command/file generator for DDP

[AyushExel]

@Laughing-q I attempted to create a workflow where the generate_ddp_command checks if the command is coming from cli. Then it creates a temp file with the training command inside if __name__ == "__main__": block and uses that file to generate the DDP command.

Note:

• the temp file is currently created in the current working dir.. We can change it to ultralytics/ once we confirm that it works
• the temp file probably needs to be deleted manually as it persists. Again, we can do it once DDP works.

I have checked until the command/file generation and the syntax in the file seems correct. Can you check if it works as supposed to with DDP also?
I have made this PR to your branch instead of committing there directly as I'm not sure if it'll work. Merge this to your branch if this works.

🛠️ PR Summary

_{Made with ❤️ by Ultralytics Actions}

🌟 Summary

Enhancement of multi-GPU training setup in YOLO training engine.

📊 Key Changes

• Updated the train method in trainer.py to pass self to generate_ddp_command.
• Created a new utility generate_ddp_file in dist.py that generates a temporary Python file for distributed training configuration.
• Modified generate_ddp_command in dist.py to handle CLI usage and utilize the new generate_ddp_file when necessary.

🎯 Purpose & Impact

• 🚀 The changes aim to streamline the setup process for distributed training across multiple GPUs.
• 🛠 By creating a temporary file with training configuration, it simplifies the command generation, particularly when using command line interfaces.
• 🤝 This update is expected to make it easier to initiate distributed training, potentially leading to more efficient and user-friendly multi-GPU training experiences for YOLO models.

[Laughing-q]

@AyushExel okay tested it and it works. Should I just merge it to DDP?

[AyushExel]

@Laughing-q yeah sure..do you see the merge option ?