EthanfelandClaude Opus 4.8 726d3cf06f fix(arch): detect all Krea 2 LoRA forms via the attention gate marker ...

The initial krea2 detection only matched the Comfy-Org native form
(blocks.N.attn.wq). Real trainer outputs use diffusers-style keys and were
misdetected: the 'krea_2' trainer (diffusion_model.transformer_blocks.N.attn
.to_q/to_gate) hit the ACE-Step branch, and the diffusers form
(transformer.transformer_blocks.N...) would hit qwen.

The marker common to ALL forms is the per-attention sigmoid GATE
(attn.to_gate / attn.gate / attn.w{q,k,v,o}) — unique among supported archs.
Detect on that, moved BEFORE the qwen/flux/ACE-Step branches. Verified vs 3
real LoRAs (official Comfy-Org rank-64, a krea_2-trained style LoRA, and a
diffusers 'softwatercolor' with text_fusion) — all -> krea2/dit. +2 tests
(ACE-Step/qwen collision guards); 284 pass. Registry 1.7.12.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

2026-06-23 22:02:34 +02:00

..

test_estimator_features.py

feat: add EstimatorFeatureExtractor for combo feature vectors

2026-04-21 14:04:25 +02:00

test_estimator_freshness.py

fix(estimator): check index version and harden meta/per_prefix parsing

2026-04-21 14:20:22 +02:00

test_estimator_integration.py

test(estimator): end-to-end integration (build → retrieve → emit)

2026-04-21 14:34:07 +02:00

test_estimator_node.py

fix: surface fresh per-prefix stats in estimator Phase 1

2026-04-21 14:44:37 +02:00

test_estimator_runtime.py

feat(estimator): LoRAEstimator runtime — retrieve, aggregate, emit tuner_data

2026-04-21 14:22:03 +02:00

test_index_builder.py

feat: index builder script for estimator k-NN

2026-04-21 14:06:33 +02:00

test_lora_optimizer.py

fix(arch): detect all Krea 2 LoRA forms via the attention gate marker

2026-06-23 22:02:34 +02:00

test_per_prefix_capture.py

feat(autotuner): batch 3 — community I/O, cache GC, selection replay, dedup

2026-06-10 15:23:23 +02:00