Add preset smoke-test scripts, including LoRA load checks (diffusers + ComfyUI)#1545
Draft
dxqb wants to merge 2 commits into
Draft
Add preset smoke-test scripts, including LoRA load checks (diffusers + ComfyUI)#1545dxqb wants to merge 2 commits into
dxqb wants to merge 2 commits into
Conversation
This was referenced Jun 20, 2026
Collaborator
Author
|
--preset-path lets a built-in preset be applied (without migration, since built-in presets are always saved in the current format) before the main config file. --config-value KEY=VALUE applies a final override after both, with dotted-path support for nested config objects - useful for scripting training runs without hand-editing a config file (e.g. overriding the output path or disabling compile per-run).
…+ ComfyUI) Trains a real LoRA per preset (run_lora_presets.sh) or embedding ( run_embedding_presets.sh) against a small shared minimal.json config, then verifies each trained LoRA loads via both a vanilla diffusers pipeline and ComfyUI's own loader (CPU-only, load-only) to catch real-world LoRA key-name mismatches that the existing sampling-based smoke tests don't exercise. Training is decoupled from the two load checks via --config-value (added in the previous commit), so any of the three stages (train, diffusers-load, comfy-load) can be rerun independently against already-saved LoRAs.
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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
Builds on #1531 (the
--preset-path/--config-valueCLI overrides).Adds
test/run_lora_presets.shandtest/run_embedding_presets.shto train every built-in preset against a small sharedtest/minimal.jsonconfig, plus two new load-only smoke tests run against each trained LoRA:test/smoke_load_diffusers.py- loads the LoRA into a vanilla diffusers pipeline viaload_lora_weights, to catch real-world LoRA key-name mismatches outside of OT's own loading code.test/smoke_load_comfy.py- same, but against ComfyUI's own loader (comfy.sd.load_lora_for_models).Both are CPU-only, load-only (no sampling). Training is decoupled from the two load checks via
--config-value(from #1531), so each of the three stages (train / diffusers-load / comfy-load) can be rerun independently against already-saved LoRAs inmodels/lora_smoke/<preset>/model.safetensors.Initial results across all built-in model types are summarized in #1544.
Test plan
pre-commit run --all-filespassesAI assistance
Early AI prototype — opened for discussion, not ready for review
Drafted by Claude