nodes/city96-ComfyUI-GGUF
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
128 Commits 4 Branches 0 Tags
bc5223b0e37e053dbec2ea5e5f52c2fd4b8f712a
Commit Graph
128 Commits
This Branch
This Branch
All Branches
Author SHA1 Message Date
City bc5223b0e3 Fix qtype logging on newer gguf-py 2025-03-16 20:14:33 +01:00
City 4aebf1bf2c Add compat mode pig 
closes #230
 2025-03-16 19:38:40 +01:00
City 5aa6e98ad7 Merge pull request #224 from city96/tokenizer_convert 
Sentencepiece tokenizer from gguf metadata for UMT5
 2025-03-14 17:53:02 +01:00
City ad69decc8d Update requirements.txt 2025-03-14 17:43:05 +01:00
City 31dffa061b Add error message for missing packages 2025-03-14 17:37:14 +01:00
City 296b0c1419 Clean up field logic 2025-03-14 17:11:47 +01:00
City d99b2d931f force dequant on umT5 to prevent OOM 
not the best solution but that tensor alone needs ~3GB of runtime memory to dequant
 2025-03-14 16:56:48 +01:00
City 01f6e3ff63 spiece tokenizer from gguf metadata 2025-03-05 04:44:53 +01:00
City 298192ed60 Wan model arch 2025-02-26 07:40:44 +01:00
City cd361cf0e3 Update README.md 2025-02-26 07:39:01 +01:00
YarvixPA 8832d37df2 Solution for patch files line endings and update for /tools README.md (#202) 
* Create convert_lines_ending.py

Solution to convert line ending from Windows (CRLF) to Unix (LF)

* Update README.md

* Rename convert_lines_ending.py to fix_lines_ending.py

* Update README.md
 2025-02-26 07:35:42 +01:00
City e024aab10d Torch compile possible workaround 
Seems to help on torch 2.6.0 with triton 3.2.0, less so on old versions.
Includes workaround for 2.0.X and other really old versions
#118
 2025-02-23 18:49:50 +01:00
City 5875c52f59 Unpin numpy version 
#196 #7
 2025-01-08 16:30:17 +01:00
City 8098c7149f Add PixArt clip type 
Also makes the map more readable
#195
 2025-01-04 21:46:23 +01:00
City 3dc384b233 More robust folder path logic 
Attempts to override the unet_gguf key with our current one if it exists + print a warning.
Should only be triggered if another custom node edits the expected key in some way.
 2024-12-26 21:42:33 +01:00
City c6ad4f2363 Account for not having quantized weights 
#188
 2024-12-22 19:00:03 +01:00
City 51fa2cbb5c Account for dequantization VRAM to avoid OOM 
#187
 2024-12-22 18:09:51 +01:00
City eaffa0a7d3 Change clip workaround for SD3 
#186
 2024-12-22 15:24:14 +01:00
City 8d1b56c5e2 Loader logic for LLaMA models 
Tokenizer shape may be mismatched depending which model was used as source (128320 VS 128256) - warning instead of error for now.
k and q attn keys need to be reshaped back to HF format (inverse of convert_hf_to_gguf.py logic)
 2024-12-20 13:39:17 +01:00
City 0a1723baae Split loader into separate file 2024-12-20 13:03:27 +01:00
City 0b65261cc4 HunyuanVideo 
Convert script needs fixing, had to use a second script for the 5D tensor, will clean up and upload it later.
 2024-12-17 07:33:42 +01:00
City 65a7c895bb Better propagation fix 
#163
 2024-12-14 00:38:11 +01:00
City 23b9817081 LTX-Video 
Whilst we wait for Hunyuan Video.
 2024-12-08 00:48:59 +01:00
City 851564739d Updated Intel Arc fix 
Provided by @simonlui in #50 as a patch
 2024-12-05 03:16:23 +01:00
City 39102df73f Stop custom tensor class propagating past dequant 
#163
 2024-12-04 16:14:54 +01:00
City 6e4790f903 Unused workaround 2024-12-04 15:54:20 +01:00
City a8b5f5d892 Less fragile patch logic 
For new ComfyUI versions. Relatively minimal, hopefully won't break with future updates
#163
 2024-12-04 02:09:01 +01:00
City 4a84328841 Add AuraFlow support 2024-11-23 06:51:29 +01:00
City 57562d54a5 Add ltxv clip type 
#162
 2024-11-23 00:10:27 +01:00
City 51c9713400 Fix sd1 not using shape fix 2024-11-21 02:24:40 +01:00
RandomGitUser321 8e898fad4c Add mochi to CLIPLoaderGGUF options (#154) 
* Add mochi to CLIPLoaderGGUF options

ComfyUI implemented it today, so this should be all that's needed to use the T5 GGUF with it

* Update nodes.py
 2024-11-06 14:04:04 +01:00
City 9833348005 Fix for text encoder folder name change 
#151
 2024-11-03 14:03:42 +01:00
City 6561064dcf Add SD3.5 links 2024-10-23 04:38:39 +02:00
City b70240508f Fix SD3 patch metadata/tensor issue 
#136
 2024-10-23 03:16:14 +02:00
City 4d7bb93397 Add initial SD3 lcpp patch + instruction 
#136
 2024-10-23 01:39:45 +02:00
City 0ef158d793 Add SD3 to loader arch list 2024-10-22 22:00:37 +02:00
City 68ad5fb214 Update README.md 2024-10-15 04:11:45 +02:00
City d090a910ac Rewrite convert.py to work properly with dev de-distill 
Original logic was triggering SDXL specific code to reshape img_in weight, causing forge not to work with it.
Tagging @blepping since this might change how the SDXL code is evaluated, but based on a quick check it should work.
 2024-10-15 03:56:54 +02:00
City d2aaeb0f13 Always init tenc on offload device 
#83
needs updated ComfyUI as well to fix it, issue was with the direct load logic in model management putting the model on the GPU before the weights were loaded, at which point the loading would overwrite the weights with the CPU ones
 2024-09-17 15:26:39 +02:00
City 454955ead3 Skip initialization for nn.Linear 
Attempt to reserve less memory by not initializing linear layers until the weights are loaded.
#102 #95
 2024-09-09 01:37:11 +02:00
City ac96699b7a Update README.md 2024-09-06 18:41:49 +02:00
City 0342f0a183 never double patch 
#98
Instead of adding/removing patches, sync with `self.patches[key]` - old patch should be GC'd. Most likely no legitimate reasons for double patching. Can simplify logic in ops.py if no issues.
 2024-09-06 18:10:02 +02:00
City b46c206dd7 double patch issue fix for lowvram 
Potential fix, lowvram weights are called via `patch_weight_to_device` on each run and it was possible to apply the same patch multiple times on subsequent generations.
 2024-09-05 05:32:01 +02:00
City 717a0e1d0b Temporarily disable instant patch logic 
Seems to break VRAM estimation and needs more involved fix with upstream ComfyUI repo.
 2024-09-03 17:00:20 +02:00
City c8923a4cd7 mmap unload fix for lowvram 2024-09-03 16:37:35 +02:00
City 6dbb4ba989 Temp fix for disable smart memory 
Model management seems to expect `unpatch_model` to put model on specified device after unpatching. Needs different way to avoid patch reload.
 2024-09-03 05:28:49 +02:00
City 69f0daffc5 Merge pull request #92 from city96/ops_refactor 
Fix LoRA reloading, ops refactor, better lowvram speed, sd.cpp compat
 2024-09-03 03:13:54 +02:00
City 929d15454d Rename ggml_forward 
Makes it clear that this is the same thing as comfy_cast_weights
 2024-09-02 03:58:26 +02:00
City 35007f2659 Refactor model loading 
Should make it more readable. Moved import under if to avoid changes to stuff in `tools` breaking normal models.
 2024-09-02 03:44:07 +02:00
City df3378da6b Detect quant 2024-09-02 03:17:16 +02:00