ci: lint autofix on pull requests#48
Open
miguelenes wants to merge 85 commits into
Open
Conversation
…eed, accelerate, and other libraries; add pytest configuration and smoke tests
…gs; refactor imports in multiple files to use updated timm library structure
…t.toml; enhance README with performance tuning details
… presets, and add new inference scripts for Hunyuan and WanVideo models
…ts, adjust deepspeed version in pyproject.toml, and optimize DataLoader settings in multiple config files
…h new inference scripts for various models, and improve README with upgrade notes and CI instructions
…E with installation instructions for AMD ROCm and CPU support, and enhance poetry.lock and pyproject.toml for improved dependency management
…lated markers in poetry.lock for improved dependency management
…l, add new CUDA presets and verification script, and update README with multi-GPU support and inference configurations
…duce .gitmodules for SimpleTuner submodule, and create AGENTS.md for project guidelines and agent workflow
…encies, enhance code clarity, and improve overall project structure
…training and inference, update .gitignore to include new data directories, and enhance documentation for fine-tuning processes
…20p, implement benchmark matrix function for improved backend performance testing, and refactor optimization functions for clarity and modularity
…atrix documentation, and implement new tests for flow tier compatibility and inference limits
…ent configuration examples, and enhance documentation for GPU training workflows
…nt variables and documentation, and streamline agent instructions for LoRA training workflows
… attention backends, and remove obsolete test files
…ation and clarity in scheduler files
…on for PrivTune, and streamline training scripts for improved clarity and usability
…ing commands, and enhance testing procedures for domain LoRA workflows
Co-authored-by: Miguel Enes <miguelenes@users.noreply.github.com>
…RA, enhance testing procedures, and update documentation for improved clarity and usability
…ns, enhance dataset handling for image-video pairs, and update documentation for improved clarity on I2V workflows
…bles, streamline access to configuration in scripts, and enhance CPU inference handling
…oints, and streamline inference options handling
…ting commands to use ruff, and streamline code for improved clarity
…D, enabling high-performance model pulls during bootstrap and updating documentation for cloud users
… I/O utilities, and enhance video frame handling across the codebase
Move marker and testpaths settings into pyproject.toml and keep conftest auto-skip logic unchanged for gpu/rocm markers.
Wire Flux Accelerate tracking to local TensorBoard, mirror Wan epoch stats in TB while keeping metrics.json, drop the unused wandb dependency, and document one metrics path per training phase.
Keep a deprecated videotuna alias for backward compatibility and document the privtune image in the README install section.
Single resolve_inference_profile() path replaces split preset/offload logic and applies presets after YAML merge so config memory_preset values take effect.
Add ADR-001 for the intentional dual training stack split and link it from AGENTS.md and related docs.
Honor gradient_checkpointing=false explicitly, add regression tests for entry scripts and in-training validation previews, and document validation artifacts in the domain runbook.
Consolidate CPU smoke tests into poetry run coverage-gate with a 33% line-coverage floor on videotuna/training/ and videotuna/utils/.
…ion. Resume loads checkpoint-* LoRA weights with fail-fast when missing, wires gradient_accumulation_steps from config, and documents validate-domain-t2v as production-ready with GPU integration coverage on both Wan 2.2 experts.
…mmit guards. Document CUDA-only Wan training on ROCm, reject undeclared MPS backends, enforce sequential CPU offload precedence when both flags are set, and narrow pre-commit to the CI smoke coverage gate.
… CSV validation. Wire adamw_bf16 to torch-optimi, pre-fetch Wan train/validate weights on cloud bootstrap, replace substring LoRA target matching, and reject misconfigured I2V dataset CSV schemas at init.
Declare torchcodec as an optional Poetry extra, prefer it in auto backend when installed with PyAV fallback, and document the install path for training.
Wire optimum-quanto as an opt-in Poetry extra so --quant-backend quanto is installable, with a clear import guard and a GPU smoke test for int4_wo on the Wan 2.2 low-VRAM preset.
…tartup log. Confirm adamw_bf16 round-trips from flux_t2i.json and log which optimizer implementation is selected at training start. Core wiring was already present in _create_optimizer (optimi.AdamW vs torch.optim.AdamW). Co-authored-by: Miguel Enes <miguelenes@users.noreply.github.com>
…ference. Bump torchao, migrate Diffusers quant configs, add fp8 preset and hardware-tier docs, and expand CPU/GPU validation coverage.
Introduce VIDEOTUNA_METRICS_BACKEND setting (tensorboard default, trackio enables dual logging alongside TensorBoard). Wire Flux Accelerate trainer to resolve log_with dynamically and log validation images to Trackio when enabled. Add optional trackio Poetry extra (~0.20.0 for huggingface-hub compatibility), resolver utilities, tests, and runbook documentation. Co-authored-by: Miguel Enes <miguelenes@users.noreply.github.com>
Resolve poetry.lock content-hash conflict by regenerating the lockfile after merging main's torchao ^0.15.0 bump with the trackio optional extra. Co-authored-by: Miguel Enes <miguelenes@users.noreply.github.com>
Add optional Trackio metrics backend for Flux training
Harden torch-optimi Flux optimizer wiring (verify + logging)
Set up Cloud dev environment (CPU profile) + AGENTS.md notes
…vements - Introduced a new `.coverage` file for coverage reporting. - Updated `pyproject.toml` to include McCabe complexity checks and per-file ignores for specific paths. - Added `fix_e501.py` script to automatically wrap long lines exceeding the specified length. - Refactored docstrings and comments across multiple files for better readability and compliance with line length standards. - Improved error handling by specifying exceptions in several try-except blocks. - Removed unused static methods and cleaned up imports in various modules. - Enhanced logging messages for clarity and consistency.
Re-run diffusers 0.36-0.38 x peft 0.17/0.18/0.19 matrix against bridge tests; confirm production combo diffusers 0.38.0 + peft 0.17.1. Document upstream dual-expert API alignment, transformers row, corrected peft 0.19 failure (torchao pin), and matrix procedure in MODEL_VERSIONS.md. Add CPU test for offline export via load_lora_adapter(prefix=None). Co-authored-by: Miguel Enes <miguelenes@users.noreply.github.com>
Audit Wan 2.2 LoRA bridge pins (diffusers/peft/accelerate/safetensors)
Restore valid string literals and expressions in wanvideo, transforms, schedulers, and utils modules that were broken by mid-string line breaks.
- Updated docstrings across multiple files for clarity and consistency. - Reformatted long lines to adhere to style guidelines. - Replaced type checks with isinstance for better practice. - Enhanced error messages for better debugging. - Removed unnecessary variable assignments and streamlined code logic. - Improved comments for better understanding of the code flow.
Add a PR-only workflow that runs poetry run format and auto-commits fixes. Include a deliberately misformatted test file to verify the flow.
Drop Poetry and full dev install; install ruff from pyproject.toml via astral-sh/ruff-action and run the same fix commands directly.
Avoid ruff-action resolving ^0.6.8 to a newer release with different lint behavior.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
.github/workflows/lint-autofix.ymlto auto-applypoetry run formaton same-repo PRs and push fixes back to the branch.Test plan
Lint autofixworkflow runs and pushes astyle: apply ruff lint and format fixescommitCPUworkflow passes lint + format-check