Repository navigation
Replace the SNOFS LoKR with the merged SNOFS Krea 2 model - #19
Merged
Merged
Conversation
Invoke 6.14.1/6.14.2 can't load the SNOFS LoKR: its Krea 2 LoRA loader has no lokr_* suffixes, so generating fails with "Unsupported lora format: dict_keys(['alpha'])" and the model list shows it as Unknown (fix is invoke-ai/InvokeAI#9424, still open). Remove the LoKR from krea-2-turbo, which is keyless and sfw again. Add snofs-krea-2: CivitAI model 2416142 "Krea 2 Raw v1.4" (bf16, 25.6 GB, SNOFS merged in), saved with a "raw" token so Invoke reads the Base variant, 52 steps / CFG 3.5, 32 GB GPU, needs the CivitAI key. Bundles Krea's Turbo LoRA for an optional 8-step setup. The Turbo merge is int8 only, which Invoke can't load. Live check (RTX A6000, $0.542/hr): Ready in 1256 s, registered as krea-2/main variant krea2_base, defaults applied, two 1024 images generated. 8/8 PASS, ~$0.33. The Turbo LoRA was registered but not used in a generation. Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
This was referenced Oct 1, 2026
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Why
Invoke 6.14.1 and 6.14.2 can't load the SNOFS LoKR bundled in #17. Its Krea 2 LoRA loader has no
lokr_*suffixes, so onlyalphais left in each layer. Generating fails withValueError: Unsupported lora format: dict_keys(['alpha']), and the model list shows the file as Unknown. The file itself is a normal ai-toolkit LoKR (256 layers ofalpha,lokr_w1,lokr_w2). The upstream fix is invoke-ai/InvokeAI#9424, still open.What
krea-2-turbo: remove the LoKR. It is keyless and not flagged nsfw again.snofs-krea-2: CivitAI model 2416142 "Krea 2 Raw v1.4" (bf16, 25.6 GB, SNOFS merged in), saved assnofs-krea-2-raw-v1.4.safetensorsso Invoke reads the Base variant. 52 steps / CFG 3.5, 32 GB GPU, needs the CivitAI key. Bundles Krea's Turbo LoRA (ungated, pinned HF revision) for an optional 8-step setup. The Turbo merge on CivitAI is int8 only, which Invoke can't load.docs/findings.md.Testing
cargo test(catalog tests) andcargo clippy -- -D warningspass.TEST_MODEL=snofs-krea-2 GENERATE=2 LOCAL_CATALOG=1, cap $0.70/hr): RTX A6000, $0.542/hr. Ready in 1256 s, registered askrea-2/mainvariantkrea2_base, defaults applied, two 1024 images generated (156 s and 135 s). 8/8 PASS, about $0.33. Instance destroyed.🤖 Generated with Claude Code