mflux pull request workflow
When to Use
- You’re about to open a PR (or want a safe sequence to do it).
Instructions
- If you run tests as part of PR hygiene, prefer fast tests first:
- Keep commits focused and messages consistent with repo history.
- If the PR changes CLI defaults, public APIs, or model behavior, check for README/example drift before opening the PR.
- Always ask for permission before pushing to the remote repository.
- If
gh isn’t available, fall back to the GitHub web UI (or stop and ask).
Pre-merge checklist (model port PRs)
Use after the core port lands and you are polishing for merge. For the full integration surfaces tick list (LoRA key formats, save routing, tokenizer edge cases, etc. learned from past closed PRs), see mflux-model-porting → Integration surfaces checklist.
Correctness
just lint and just test-fast- Slow golden tests for the new model:
MFLUX_PRESERVE_TEST_OUTPUT=1 uv run pytest tests/image_generation/test_generate_image_<model>.py -m slow -v
- Optional but high-signal: diffusers side-by-side + latent injection (
mflux-debugging, mflux-manual-testing)
Cross-model diff audit
List files changed outside src/mflux/models/<model>/:
| Category |
Expected |
pyproject.toml, cli/defaults/defaults.py, ModelConfig, mflux-save routing |
Required wiring |
README.md table + attribution |
Required |
Training runner.py, example JSON, .gitignore JSON exceptions |
If training supported |
| Shared VAE/callback/training one-liners |
Only if required; document blast radius in PR |
Personal .gitignore, unrelated formatting |
Remove |
Verify quantized README disk claims with measurement:
du -sh ~/.cache/huggingface/hub/models--<org>--<Model>*
mflux-save --model <alias> --quantize 8 --path /tmp/model-q8 && du -sh /tmp/model-q8
Docs / examples
- Model README matches a recent port (e.g. Flux2): hero image, turbo + base CLI, feature section, disk warning, Notes, Training.
- Main
README.md model table row (correct release date).
- Showcase asset if other models have one (
src/mflux/assets/; may need git add -f when *.jpg is gitignored).
PR description callouts
- Shared code touched and why (shared VAE, callbacks, training runner, etc.).
- Reference pipeline features intentionally not ported (optional preprocessors, extra encoders, components omitted from mflux weight downloads).
- Known non-parity with diffusers (RNG, sigma schedule, optional modules) if golden tests lock mflux-native sampling.
1---2name: mflux-pr3description: Make a clean PR in mflux (inspect diff, quick verification, commit, push, open PR) using repo conventions.4---5# mflux pull request workflow67## When to Use89- You’re about to open a PR (or want a safe sequence to do it).1011## Instructions1213- If you run tests as part of PR hygiene, prefer fast tests first:14 - `just test-fast`15- Keep commits focused and messages consistent with repo history.16- If the PR changes CLI defaults, public APIs, or model behavior, check for README/example drift before opening the PR.17- **Always ask for permission** before pushing to the remote repository.18- If `gh` isn’t available, fall back to the GitHub web UI (or stop and ask).1920## Pre-merge checklist (model port PRs)2122Use after the core port lands and you are polishing for merge. For the full **integration surfaces** tick list (LoRA key formats, save routing, tokenizer edge cases, etc. learned from past closed PRs), see `mflux-model-porting` → *Integration surfaces checklist*.2324### Correctness25261. `just lint` and `just test-fast`272. Slow golden tests for the new model:28 ```sh29 MFLUX_PRESERVE_TEST_OUTPUT=1 uv run pytest tests/image_generation/test_generate_image_<model>.py -m slow -v30 ```313. Optional but high-signal: diffusers side-by-side + latent injection (`mflux-debugging`, `mflux-manual-testing`)3233### Cross-model diff audit3435List files changed outside `src/mflux/models/<model>/`:3637| Category | Expected |38|---|---|39| `pyproject.toml`, `cli/defaults/defaults.py`, `ModelConfig`, `mflux-save` routing | Required wiring |40| `README.md` table + attribution | Required |41| Training `runner.py`, example JSON, `.gitignore` JSON exceptions | If training supported |42| Shared VAE/callback/training one-liners | Only if required; document blast radius in PR |43| Personal `.gitignore`, unrelated formatting | **Remove** |4445Verify quantized README disk claims with measurement:46```sh47du -sh ~/.cache/huggingface/hub/models--<org>--<Model>*48mflux-save --model <alias> --quantize 8 --path /tmp/model-q8 && du -sh /tmp/model-q849```5051### Docs / examples5253- Model README matches a recent port (e.g. Flux2): hero image, turbo + base CLI, feature section, disk warning, Notes, Training.54- Main `README.md` model table row (correct release date).55- Showcase asset if other models have one (`src/mflux/assets/`; may need `git add -f` when `*.jpg` is gitignored).5657### PR description callouts5859- Shared code touched and why (shared VAE, callbacks, training runner, etc.).60- Reference pipeline features intentionally **not** ported (optional preprocessors, extra encoders, components omitted from mflux weight downloads).61- Known non-parity with diffusers (RNG, sigma schedule, optional modules) if golden tests lock mflux-native sampling.62
mflux-community