Claude
919097c732
The H3 run trains with a loop of our own (LoRA on the attention and feed-forward projections of the 50 blocks, per-block checkpoint, uniform noise levels, musubi-tuner's approach) while ComfyUI carries the batches. Every Log record is flushed to disk as it is said and reloaded on start; the run writes each stage and step to it. ComfyUI's own offloading toggle is ignored by the new loop. UNTESTED on the real weights. Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01MpnF7GwNGMbngU11wjym7P
53 lines
1.1 KiB
Plaintext
53 lines
1.1 KiB
Plaintext
.testvenv/
|
|
|
|
# Python
|
|
__pycache__/
|
|
*.py[cod]
|
|
*.egg-info/
|
|
.pytest_cache/
|
|
|
|
# OS / editor
|
|
.DS_Store
|
|
|
|
# Local tooling / agent config
|
|
.claude/
|
|
.cursor/
|
|
|
|
# Runtime artifacts (generated refinement stores, attn maps, etc.)
|
|
refinements/
|
|
|
|
# Large local research docs
|
|
*.pdf
|
|
|
|
# Never track saved ComfyUI workflow dumps — they carry whatever was typed in the graph.
|
|
movie_editor/backend/reference/*.workflow.json
|
|
|
|
# Dev-only: the app ships as plain ES modules, users never run npm.
|
|
node_modules/
|
|
|
|
# Playwright: browsers live in a shared cache, results and traces are per-run
|
|
app/test-results/
|
|
app/playwright-report/
|
|
|
|
# User projects, and what they imported — their content, not ours. Rooted
|
|
# (leading slash): a bare "media/" also matches modules/media/, this app's own
|
|
# source code for loading media into the graph.
|
|
/projects/
|
|
/media/
|
|
/training/
|
|
/shortcuts.json
|
|
/markers.json
|
|
/quarantine.json
|
|
/addons.json
|
|
/shortcut_categories.json
|
|
/shortcut_revolver.json
|
|
/prompt_composer_memory.json
|
|
|
|
# Taste keys: what your ratings taught, per rental
|
|
/taste/
|
|
/block_influence.enabled
|
|
/block_influence.latest.json
|
|
/phrase_probe.enabled
|
|
/phrase_probe.latest.json
|
|
/log.jsonl
|