Skip to content

docs(diffusion RL v0.1): add quick start, OCR example, accuracy guide, and roadmap#23

Draft
zhihengy wants to merge 3 commits into
diffusion_RL_v0.1from
docs/diffusion_RL_v0.1_zhihengy
Draft

docs(diffusion RL v0.1): add quick start, OCR example, accuracy guide, and roadmap#23
zhihengy wants to merge 3 commits into
diffusion_RL_v0.1from
docs/diffusion_RL_v0.1_zhihengy

Conversation

@zhihengy
Copy link
Copy Markdown
Collaborator

@zhihengy zhihengy commented May 5, 2026

Summary

This draft PR adds the first batch of documentation for the diffusion RL v0.1 branch:

  • Adds a quick start guide for environment setup.
  • Adds task-level dependency documentation, including OCR reward dependencies.
  • Adds a runnable Qwen-Image OCR example under examples.
  • Adds an advanced note on rollout parallel accuracy.
  • Adds an external-facing diffusion RL roadmap.

Validation

  • Ran git diff --check origin/diffusion_RL_v0.1..HEAD.
  • Confirmed the local branch is synced with origin/docs/diffusion_RL_v0.1_zhihengy before opening this PR.

Notes

This is intentionally opened as a draft so the same PR can continue to receive follow-up documentation updates without creating duplicate PRs.

n_samples_per_prompt = 16 samples per prompt
samples_per_rollout = 16 * 16 = 256 samples
num_steps_per_rollout = 2 optimizer steps
global_batch_size = 256 / 2 = 128 samples per optimizer step
Copy link
Copy Markdown
Owner

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

we can refer to "batchsizes in miles-d" in docs here

Comment thread docs/roadmap.md
@@ -0,0 +1,56 @@
# Roadmap

This roadmap describes planned directions for miles-diffusion. It is intended
Copy link
Copy Markdown
Owner

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

maybe we should also add some of our plan about stability

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants