[chatgpt] Detached PPO Training

[CsRic]

📌 Checklist before creating the PR

• I have created an issue for this PR for traceability
• The title follows the standard format: [doc/gemini/tensor/...]: A concise description
• I have added relevant tags if possible for us to better distinguish different PRs

🚨 Issue number

Link this PR to your issue with words like fixed to automatically close the linked issue upon merge

e.g. fixed #1234, closed #1234, resolved #1234

📝 What does this PR do?

Summarize your work here.
if you have any plots/diagrams/screenshots/tables, please attach them here.

'Detached' PPO training means that the experience makers and trainers are splitted into different nodes for asynchronous training. Models are not shared.

1. Propose several classes for 'detached' manner: ExperienceMakerHolder, DetachedReplayBuffer, DetachedPPOTrainer
2. Implement Ray for Detached workflow structure
3. Examples: 1m1t. 1m2t, 2m1t, 2m2t.
4. Supported Strategies: Naive, DDP.
5. Won't affect present code.

Known issues:
1. Cannot detect cuda device on each worker. fixed 20230324
2. Cannot run with Colossal strategy.
3. correctness of 1m1t.py example. fixed 20230324

TODO:

• Implement parameter update from trainer to experience maker
• fix issues
• Add TP strategy for experience maker and trainer.
• Support multiple nodes.

💥 Checklist before requesting a review

• I have linked my PR to an issue (instruction)
• My issue clearly describes the problem/feature/proposal, with diagrams/charts/table/code if possible
• I have performed a self-review of my code
• I have added thorough tests.
• I have added docstrings for all the functions/methods I implemented

⭐️ Do you enjoy contributing to Colossal-AI?

• 🌝 Yes, I do.
• 🌚 No, I don't.

Tell us more if you don't enjoy contributing to Colossal-AI.

[binmakeswell]

Hi @CsRic Thanks for your contribution, but there is a conflict in this PR. Could you please solve them first? Thanks.

[ver217]

Can you move all files to coati/ray?

[CsRic]

Can you move all files to coati/ray?

Done