Commit Graph

Select branches

Hide Pull Requests

master

Mono Color

• 540bc9f898 Merge pull request #3 from marduk191/claude/optimize-image-embeddings-01Pm2Lp1Ghm6QaYtseCKJcGZ master Nathan aka SoCalGuitarist 2025-12-17 18:11:18 -08:00
• c30c34b244 fix: Move thread count setting to module load time Claude 2025-12-12 19:42:19 +00:00
• 5e9a042c16 feat: Make thread count user-configurable via dropdown Claude 2025-12-12 19:33:08 +00:00
• 45c0fd5b79 perf: Set PyTorch thread count to 8 for optimal CPU parallelism Claude 2025-12-12 19:31:27 +00:00
• 078991cc03 fix: Add transformers 4.56.2 compatibility patch for slow generation Claude 2025-12-11 11:00:44 +00:00
• 403936d64a revert: Remove vision prefill optimization that caused errors Claude 2025-12-11 05:00:30 +00:00
• e46e6c76bb perf: Pre-process vision tokens once to fix low GPU utilization Claude 2025-12-11 04:53:07 +00:00
• 529aa87c60 revert: Remove backend settings that degraded performance Claude 2025-12-11 03:54:43 +00:00
• 444a3f3be7 fix: Remove torch.compile and force optimal SDPA backends Claude 2025-12-11 03:38:45 +00:00
• f629c2278b perf: Add torch.compile for 2-3x generation speedup Claude 2025-12-11 01:36:39 +00:00
• c3cc6b8cdb fix: Add CUDA state reset to prevent random performance degradation Claude 2025-12-11 01:32:12 +00:00
• e51ad944d0 feat: Add detailed device verification and stall detection diagnostics Claude 2025-12-11 01:13:35 +00:00
• 1e01db3ee3 feat: Add input token count and GPU memory diagnostics Claude 2025-12-11 01:07:49 +00:00
• 87ae4e060c feat: Add comprehensive performance diagnostics and warnings Claude 2025-12-11 00:50:49 +00:00
• 51f9048931 feat: Add configurable image resolution and timing diagnostics for faster image embedding Claude 2025-12-11 00:31:42 +00:00
• 320284b9cd Clean up requirements.txt - remove dependencies already provided by ComfyUI SanDiegoDude 2025-12-09 12:17:10 -08:00
• 2ebace5986 Merge pull request #1 from EnragedAntelope/master Nathan aka SoCalGuitarist 2025-12-05 12:56:05 -08:00
• a8cd3e0301 fix: Remove duplicate repetition_penalty parameters from merge conflict EnragedAntelope 2025-12-05 14:53:00 -05:00
• da10ce4591 Merge pull request #1 from EnragedAntelope/claude/implement-fixes-enhancements-01Cxt5wVVQY4yqXMYnr3Q8ax EnragedAntelope 2025-12-05 14:47:47 -05:00
• a4ac4603d0 Merge branch 'master' into claude/implement-fixes-enhancements-01Cxt5wVVQY4yqXMYnr3Q8ax EnragedAntelope 2025-12-05 14:47:24 -05:00
• d3291f6e87 docs: Add performance and attention mode details to README and a detailed performance optimization session summary to context. EnragedAntelope 2025-12-05 14:42:15 -05:00
• 600340d91d feat: Refactor Flash Attention 2 compatibility to optimization check, improve device mapping for model loading, and enhance attention mode selection logic. EnragedAntelope 2025-12-05 14:30:37 -05:00
• 844494fa37 simp2 EnragedAntelope 2025-12-05 13:47:40 -05:00
• 749d78fa45 SIMPLIFIED for maximum performance EnragedAntelope 2025-12-05 13:46:19 -05:00
• ecf45acdde feat: Enhance attention mode selection by detecting GPU compatibility for Flash Attention 2 and defaulting to SDPA when incompatible, unavailable, or for auto-selection. EnragedAntelope 2025-12-05 13:31:48 -05:00
• 3a542036c8 fix dict issue EnragedAntelope 2025-12-05 13:01:33 -05:00
• 9cd8d0a771 perf: Improve non-quantized model loading with device_map and optimize input tensor transfer to GPU. EnragedAntelope 2025-12-05 12:27:18 -05:00
• b5f8ecf9a4 refactor: Remove torch.compile attempts and ensure models are set to evaluation mode after loading. EnragedAntelope 2025-12-05 12:02:16 -05:00
• 2b0155a826 feat: Enhance model loading and generation performance by explicitly moving non-quantized models to GPU, integrating torch.compile, and adding CUDA synchronization for accurate timing. EnragedAntelope 2025-12-05 11:30:16 -05:00
• a668a2c7ad feat: Add quantization support check and refine model loading with conditional torch_dtype, device_map, and BetterTransformer integration. EnragedAntelope 2025-12-05 11:09:27 -05:00
• 8bea6cfaa3 Add repetition_penalty parameter to prevent output loops SanDiegoDude 2025-12-05 07:32:30 -08:00
• a552e89003 Fix 4bit quantization performance + ComfyUI v3 compliance Claude 2025-12-05 15:32:03 +00:00
• abaeb4aa8e COMPREHENSIVE PERFORMANCE FIXES: 4bit optimization + logging + tensor ops Claude 2025-12-05 14:49:38 +00:00
• ceaece4926 Fix 4bit quantization performance + ComfyUI v3 compliance Claude 2025-12-05 14:38:40 +00:00
• 628694c245 CRITICAL FIX: Add torch_dtype to ALL model loading paths Claude 2025-12-05 14:23:36 +00:00
• b1e8146551 Implement comprehensive fixes and enhancements to QwenVL nodes Claude 2025-12-05 13:57:05 +00:00
• 034e806f2d Fix 4-bit quantization running on CPU instead of GPU SanDiegoDude 2025-12-04 20:19:37 -08:00
• 7b8425e6cd Add custom models support, multi-image input, bypass mode, and increased token limit SanDiegoDude 2025-12-04 18:10:50 -08:00
• 56f28865ff Bump version from 2.1.1 to 2.1.3 Alex Cong 2025-11-29 15:16:15 -05:00
• 51c5227829 Merge pull request #29 from alexcong/issue-28 Alex Cong 2025-11-29 15:12:06 -05:00
• 8c6cd79407 Allow system-only prompts Alex Cong 2025-11-29 14:00:36 -05:00
• 29bc1de537 Merge pull request #27 from silveroxides/patch-1 Alex Cong 2025-11-12 22:17:22 -05:00
• 6bc9f0de40 Remove deprecated local_dir_use_symlinks Silver 2025-11-12 20:13:24 +01:00
• da5056b7f0 minor project ver update Alex Cong 2025-11-08 15:48:04 -05:00
• 8017304cbd add github action Alex Cong 2025-11-08 15:46:24 -05:00
• 7dfb1dab3c update project version and description Alex Cong 2025-11-08 15:27:07 -05:00
• aab3d4990d Merge pull request #26 from alexcong/clearvram Alex Cong 2025-11-08 15:20:54 -05:00
• 8b7388eb6f Update agent documentation Alex Cong 2025-11-08 13:47:03 -05:00
• 0c05a660f5 Add VRAM cleanup helpers and tests Alex Cong 2025-11-08 11:53:23 -05:00
• f29a4be1e4 Add new Qwen3 models to the model list Alex Cong 2025-10-21 13:20:15 -04:00
• 028ea49e51 update readme Alex Cong 2025-10-20 16:07:24 -04:00
• 50b700bde1 Merge pull request #24 from alexcong/qwen3vl Alex Cong 2025-10-20 15:58:46 -04:00
• b871a9d79a update node names Alex Cong 2025-10-20 15:57:01 -04:00
• a3be25f980 typo fix Alex Cong 2025-10-20 15:32:31 -04:00
• d0e640dcbe add more models, update transformers verson Alex Cong 2025-10-20 15:14:49 -04:00
• 31429837ca Add Qwen3VL 4B and 8B Thinking model support Alex CongandClaude 2025-10-20 14:05:02 -04:00
• 841661f2e7 Merge pull request #18 from throttlekitty/patch-1 Alex Cong 2025-05-03 17:43:37 -04:00
• e9681ca429 Add SkyCaptionerV1, a fine tune for video prompt generation. throttlekitty 2025-04-22 11:06:33 -06:00
• 00e8083e4b Merge pull request #14 from ComfyNodePRs/licence-update Alex Cong 2025-03-14 21:11:06 -04:00
• b12bef729d Merge pull request #13 from Nojahhh/Nojahhh-patch-1 Alex Cong 2025-03-14 21:10:35 -04:00
• ff68c97c26 chore(licence-update): Update PyProject Toml - License snomiao 2025-03-08 17:25:35 +00:00
• 21b4eb01db Update nodes.py Johan Mellin 2025-03-04 02:45:24 +01:00
• 3e9ee34132 Updated code to align with qwen_vl_utils package Johan Mellin 2025-03-04 00:43:38 +01:00
• 1a045d873d Merge pull request #9 from civen-cn/master Alex Cong 2025-02-15 14:38:11 -05:00
• 1107f8e4b1 Merge remote-tracking branch 'origin/master' civen-cn 2025-02-15 22:36:53 +08:00
• 8afbd90eb3 2.5 VL for video civen-cn 2025-01-27 18:30:11 -05:00
• 720daf4733 Update README.md Alex Cong 2025-01-27 18:30:11 -05:00
• 64cbfecd5a Update README.md Alex Cong 2025-01-27 18:27:45 -05:00
• 04f168eb64 Update node name to qwen2.5vl Alex Cong 2025-01-27 18:18:30 -05:00
• 3fdaee97b9 remove Qwen2_VLForConditionalGeneration, alexcong 2025-01-27 14:42:18 -05:00
• c5e42ecb00 remove qwen2vl models alexcong 2025-01-27 14:41:40 -05:00
• d652daa28e Update to Qwen2_5_VLForConditionalGeneration Alex Cong 2025-01-27 14:28:57 -05:00
• e3c39b8ca1 Add Qwen2.5-VL-7B-Instruct and Qwen2.5-VL-3B-Instruct Alex Cong 2025-01-27 14:26:06 -05:00
• 08bbd1c1db Update min/max pixels settings Alex Cong 2024-12-28 23:17:34 +00:00
• f19ef03e93 Merge pull request #4 from alexcong/dev Alex Cong 2024-12-28 22:56:27 +00:00
• c66a9d456b remove text input check Alex Cong 2024-12-28 17:22:20 -05:00
• a181f98f18 add min_pixels and max_pixels to qwen2vl Alex Cong 2024-12-28 17:19:05 -05:00
• 4e2cb62b9f update readme Alex Cong 2024-09-22 17:15:43 +00:00
• c10cf73d24 fix device Alex Cong 2024-09-22 17:01:29 +00:00
• 5f79217a4b add qwen2.5 text generation Alex Cong 2024-09-22 16:29:48 +00:00
• f7909db55f Create LICENSE Alex Cong 2024-09-09 17:05:27 -04:00
• 78679a3aee Update README.md Alex Cong 2024-09-09 13:16:44 -04:00
• 4e17e0e525 update readme alexcong 2024-09-09 09:50:46 -04:00
• 4175f662e8 update workflow picture alexcong 2024-09-09 09:46:43 -04:00
• 238d50ed69 update readme, add workflow json alexcong 2024-09-09 09:39:21 -04:00
• 4d698e28a4 update readme alexcong 2024-09-09 09:23:56 -04:00
• 85e368d0d0 Merge branch 'master' of https://github.com/alexcong/ComfyUI_QwenVL alexcong 2024-09-09 09:23:40 -04:00
• 718fed7023 update readme alexcong 2024-09-09 09:22:13 -04:00
• 1509df5041 Update README.md Alex Cong 2024-09-08 21:59:10 -04:00
• 8282a71f5f fix tensor check alexcong 2024-09-08 21:43:26 -04:00
• ca60a147d4 fix tensor to pil conversion alexcong 2024-09-08 21:39:25 -04:00
• bd4c69d77a image type fix alexcong 2024-09-08 21:26:01 -04:00
• ba622536b0 fix error alexcong 2024-09-08 20:54:49 -04:00
• 2225bff64d add pyproject alexcong 2024-09-08 20:40:27 -04:00
• bab56ffb13 update requirements alexcong 2024-09-08 20:40:15 -04:00
• dbbb2724ec add qwen2vl node alexcong 2024-09-08 19:52:33 -04:00