Commit Graph

Select branches

Hide Pull Requests

main

Mono Color

• 85409d588a Implement comprehensive image format guarantee system main Junst 2025-09-24 19:23:02 +09:00
• 45223df3d0 Implement to_preview_image function for robust tensor conversion Junst 2025-09-24 19:17:16 +09:00
• 7d3e2e52fe Implement comprehensive 2D tensor conversion and error handling Junst 2025-09-24 19:08:39 +09:00
• 962f4841fc Fix undefined target_h and target_w variables Junst 2025-09-24 18:55:17 +09:00
• f1d059b1fe Fix 3D numpy array issue in _convert_to_comfyui_format Junst 2025-09-24 18:51:10 +09:00
• 804e4d2c66 Improve 2D tensor validation and shape checking Junst 2025-09-24 18:47:48 +09:00
• 275ae35822 Fix numpy array handling in _convert_to_comfyui_format Junst 2025-09-24 18:42:58 +09:00
• 570255e0ab Fix numpy array handling in visualization Junst 2025-09-24 18:36:15 +09:00
• 999b2c448e Fix 3D tensor issue in concept map creation Junst 2025-09-24 17:53:31 +09:00
• a040e5bf11 Fix image visualization and numpy array handling Junst 2025-09-24 17:49:17 +09:00
• 1ee4ecfe88 Fix concept_list string to list conversion Junst 2025-09-24 17:41:20 +09:00
• bbb3b3ed2c Fix concept map creation logic and remove duplicate processing Junst 2025-09-24 17:37:47 +09:00
• 650120b474 Fix concept map creation and PIL compatibility Junst 2025-09-24 17:33:52 +09:00
• c6269162a5 Fix PIL image compatibility and tensor shape issues Junst 2025-09-24 17:27:36 +09:00
• 80e07c6c36 Fix concept map creation to generate proper concept maps Junst 2025-09-24 17:20:36 +09:00
• e784ef2fca Fix spatial dimension mismatch in concept map creation Junst 2025-09-24 17:15:03 +09:00
• 3499ca8a0a Fix tuple unpacking error in apply_model return values Junst 2025-09-24 17:09:42 +09:00
• f5797fdca2 Fix device mismatch error in concept embeddings Junst 2025-09-24 17:04:48 +09:00
• 5625021c12 Fix context and x tensor dimension mismatch with linear projection Junst 2025-09-24 17:00:20 +09:00
• fbc90b8720 Fix Flux model 3D tensor requirement Junst 2025-09-24 16:54:42 +09:00
• bf2a41b2ca Fix context and x tensor dimension mismatch Junst 2025-09-24 16:49:29 +09:00
• 78795e5de4 Fix 3D to 4D context tensor conversion for Flux model Junst 2025-09-24 16:45:09 +09:00
• ae7c51af1b Fix tensor dimension mismatch for context embeddings Junst 2025-09-24 16:38:59 +09:00
• ebe24f6ad3 Fix tensor size mismatch in Flux model inputs Junst 2025-09-24 16:28:28 +09:00
• 9b0aa20e90 Fix context parameter duplication and remove debug code Junst 2025-09-24 16:24:43 +09:00
• 3770b08735 Fix ModelPatcher callable error in concept_attention.py Junst 2025-09-24 16:20:43 +09:00
• d6b546b653 Clean up and optimize workflow Junst 2025-09-24 16:13:50 +09:00
• 88b824b0ea Fix DualCLIPLoader outputs in workflow Junst 2025-09-24 16:05:44 +09:00
• 0382c3945e Fix workflow widgets_values for ConceptAttentionNode Junst 2025-09-24 16:02:32 +09:00
• 15e737d94c Fix workflow validation by adding missing inputs Junst 2025-09-24 16:00:11 +09:00
• 627340945c Add ComfyUI node attributes for proper validation Junst 2025-09-24 15:58:57 +09:00
• 393efbeb15 Fix TITLE attribute error in __init__.py Junst 2025-09-24 15:57:14 +09:00
• 792bd5fc50 Fix relative import for ComfyUI compatibility Junst 2025-09-24 15:56:10 +09:00
• 4ccaf399e0 Simplify ConceptAttention based on original structure Junst 2025-09-24 15:54:47 +09:00
• ad6c2dc781 Remove all fallback and mock mechanisms as requested Junst 2025-09-24 15:42:03 +09:00
• ca8fdd904d Replace problematic model forward pass with fallback attention generation Junst 2025-09-24 15:38:17 +09:00
• 780b819553 Fix image shape handling in _try_model_forward_pass Junst 2025-09-24 15:34:30 +09:00
• ba9492e431 Fix image attribute error with dummy image fallback Junst 2025-09-24 15:28:35 +09:00
• 3eafd03e7b Fix model and image variable references in _try_model_forward_pass Junst 2025-09-24 15:23:40 +09:00
• 811cf27147 Fix ModelPatcher model extraction for DiffusionModelLoader Junst 2025-09-24 15:17:36 +09:00
• c781105b94 Fix DiffusionModelLoader model structure handling Junst 2025-09-24 15:13:53 +09:00
• b33a174c78 Add comprehensive model debugging for forward pass Junst 2025-09-24 15:08:19 +09:00
• 096c4a055c Add detailed error logging for model forward pass Junst 2025-09-24 15:04:51 +09:00
• 6da6b6f9b3 Add model forward pass to trigger attention hooks Junst 2025-09-24 15:01:08 +09:00
• 7282bdf62f Fix indentation error in concept_attention.py Junst 2025-09-24 14:52:43 +09:00
• cfadc10246 Fix ConceptDiffusion for Flux dual CLIP architecture Junst 2025-09-24 14:46:38 +09:00
• bf9116b43f Implement ComfyUI CLIP encode method integration Junst 2025-09-24 14:33:31 +09:00
• 6d647b2e43 Add comprehensive CLIP structure analysis for text encoder extraction Junst 2025-09-24 14:29:42 +09:00
• 5b721f0fdc Implement ComfyUI Flux-specific attention capture Junst 2025-09-24 14:25:04 +09:00
• e1eb17c0c5 Add debug logging for clip object structure Junst 2025-09-24 14:17:53 +09:00
• d3832774ae Fix text encoder and tokenizer extraction from clip Junst 2025-09-24 14:15:44 +09:00
• e222ca9869 Fix syntax error - add missing except block Junst 2025-09-24 14:11:51 +09:00
• 832af72129 Remove ALL fallback mechanisms - require real attention capture Junst 2025-09-24 14:09:01 +09:00
• 2cfc9be7d8 Fix image shape extraction for proper dimensions Junst 2025-09-24 14:03:15 +09:00
• eca5692914 Fix image shape handling and concept map conversion Junst 2025-09-24 14:01:12 +09:00
• 59be38dcab Fix matmul operation for concept attention computation Junst 2025-09-24 13:58:45 +09:00
• c7a0589ad4 Fix attention module detection and shape handling issues Junst 2025-09-24 13:56:12 +09:00
• 85d0f4d015 Fix forward method replacement and dtype conversion issues Junst 2025-09-24 13:53:11 +09:00
• ab48d4d5e5 Fix import error and improve attention capture with ComfyUI-Attention-Distillation method Junst 2025-09-24 13:48:46 +09:00
• 19e7b7c124 Implement direct forward method replacement for attention capture Junst 2025-09-23 21:16:48 +09:00
• 08a77e81fe Remove unnecessary concept_attention_clean.py file Junst 2025-09-23 21:12:30 +09:00
• f9bf2281f7 Fix patches_replace system with correct Flux model parameters Junst 2025-09-23 21:11:02 +09:00
• 610bf49aa8 Refactor to use ComfyUI patches_replace system instead of hooks Junst 2025-09-23 21:07:07 +09:00
• 5af797cf21 Implement aggressive attention capture with multiple fallback strategies Junst 2025-09-23 21:00:36 +09:00
• 16348e233c Fix concept embedding extraction and enhance attention capture Junst 2025-09-23 20:57:26 +09:00
• 7bbc1495a6 Fix iterator error and improve attention capture Junst 2025-09-23 20:53:24 +09:00
• 25284ba92c Fix attention_outputs access in direct attention execution Junst 2025-09-23 20:44:37 +09:00
• dd2b2e0d8a Fix attention_outputs attribute access issue Junst 2025-09-23 20:41:17 +09:00
• 2ea5df78d5 Add direct attention output capture and enhanced hook debugging Junst 2025-09-23 20:36:24 +09:00
• 1f8dc90cf8 Add direct attention module execution to force hook triggering Junst 2025-09-23 20:32:43 +09:00
• da4ddaea2e Implement proper DoubleStreamBlock.forward() execution based on ComfyUI analysis Junst 2025-09-23 20:26:15 +09:00
• 7af2c5bbc9 Add mock concept maps fallback and improve ConceptAttention compatibility Junst 2025-09-23 20:11:54 +09:00
• 61f68ba359 Add comprehensive attention module execution with multiple input shapes Junst 2025-09-23 20:03:57 +09:00
• f101eeab4a Add full block execution to trigger hooks Junst 2025-09-23 20:00:57 +09:00
• 19fc380b08 Implement direct attention module execution to trigger hooks Junst 2025-09-23 19:57:56 +09:00
• d3132fd585 Fix IndentationError in concept_attention.py Junst 2025-09-23 19:54:50 +09:00
• b26392df82 Implement direct Flux model forward pass to trigger hooks Junst 2025-09-23 19:50:53 +09:00
• 9d0f0952d5 Remove mock data and fix quantile dtype error Junst 2025-09-23 19:47:22 +09:00
• 4486b52904 Fix mock data tensor creation and improve SelfAttention triggering Junst 2025-09-23 19:43:45 +09:00
• 505daa562c Focus on actual attention outputs instead of Linear layers Junst 2025-09-23 19:39:49 +09:00
• cf8e46ad28 Fix attention layer selection for captured Linear outputs Junst 2025-09-23 19:37:24 +09:00
• b801082610 Fix third IndentationError in concept_attention.py Junst 2025-09-23 19:33:04 +09:00
• b64279ceeb Fix second IndentationError in concept_attention.py Junst 2025-09-23 19:32:04 +09:00
• 207c648dab Fix IndentationError in concept_attention.py Junst 2025-09-23 19:30:48 +09:00
• 7e77cf7a23 Improve hook triggering and add realistic mock data Junst 2025-09-23 19:28:49 +09:00
• af64908960 Integrate original ConceptAttention approach with ComfyUI Flux Junst 2025-09-22 14:53:44 +09:00
• f406afd060 Improve hook triggering and error handling Junst 2025-09-22 13:58:51 +09:00
• cc6ef2484a Clean up backup file and finalize concept_attention.py Junst 2025-09-22 13:53:08 +09:00
• 3a0d0dd98d Fix SyntaxError: duplicate except blocks and indentation Junst 2025-09-22 13:44:43 +09:00
• 84af9f40f8 Clean up excessive debugging logs Junst 2025-09-22 13:42:46 +09:00
• ac02db524e Implement original ConceptAttention approach with cross-attention Junst 2025-09-22 13:33:59 +09:00
• 61a983962d Fix tensor dimension mismatch in embedding processing Junst 2025-09-22 13:26:02 +09:00
• 9296033cec Fix spatial dimension processing and similarity computation Junst 2025-09-22 13:22:18 +09:00
• 764edff4e6 Fix spatial dimension issue and add concept labeling Junst 2025-09-22 13:18:00 +09:00
• 21ae900e90 Fix final device mismatch in ConceptSegmentationNode Junst 2025-09-22 11:39:21 +09:00
• 987359d653 Fix final device mismatch in ConceptAttentionVisualizerNode Junst 2025-09-22 11:27:19 +09:00
• 1e39213ea8 Fix final 'list' object is not callable error in VisualizerNode Junst 2025-09-22 11:15:52 +09:00
• 298a2ccef6 Fix device mismatch and improve error handling Junst 2025-09-22 11:11:16 +09:00
• b05841b644 Fix 'list' object is not callable error in ConceptMaps Junst 2025-09-22 11:06:21 +09:00
• 9d97ad38fa Add comprehensive debugging for concept_maps data flow Junst 2025-09-22 11:01:28 +09:00