Commit Graph

  • 08f92d55e9 Partial model shift support. comfyanonymous 2024-08-08 03:27:37 -04:00
  • 8115d8cce9 Add Flux fp16 support hack. comfyanonymous 2024-08-07 15:08:39 -04:00
  • 6969fc9ba4 Make supported_dtypes a priority list. comfyanonymous 2024-08-07 15:00:06 -04:00
  • cb7c4b4be3 Workaround for lora OOM on lowvram mode. comfyanonymous 2024-08-07 14:30:54 -04:00
  • 1208863eca Fix "Comfy" lora keys. comfyanonymous 2024-08-07 13:49:31 -04:00
  • e1c528196e Fix bundled embed. comfyanonymous 2024-08-07 13:30:45 -04:00
  • 17030fd4c0 Support for "Comfy" lora format. comfyanonymous 2024-08-07 13:18:32 -04:00
  • c19dcd362f Controlnet code refactor. comfyanonymous 2024-08-07 12:59:28 -04:00
  • 1c08bf35b4 Support format for embeddings bundled in loras. comfyanonymous 2024-08-07 03:45:25 -04:00
  • 2a02546e20 Add type hints to folder_paths.py (#4191) PhilWun 2024-08-07 03:59:34 +02:00
  • b334605a66 Fix OOMs happening in some cases. comfyanonymous 2024-08-06 13:27:48 -04:00
  • de17a9755e Unload all models if there's an OOM error. comfyanonymous 2024-08-06 03:30:28 -04:00
  • c14ac98fed Unload models and load them back in lowvram mode no free vram. comfyanonymous 2024-08-06 03:22:39 -04:00
  • 2894511893 Clone taesd with depth of 1 to reduce download size. (#4232) Robin Huang 2024-08-05 22:46:09 -07:00
  • f3bc40223a Add format metadata to CLIP save to make compatible with diffusers safetensors loading (#4233) Silver 2024-08-06 07:45:24 +02:00
  • 841e74ac40 Change browser test CI python to 3.8 (#4234) Chenlei Hu 2024-08-06 01:27:28 -04:00
  • 2d75df45e6 Flux tweak memory usage. comfyanonymous 2024-08-05 21:58:28 -04:00
  • 1abc9c8703 Stable release uses cached dependencies (#4231) Robin Huang 2024-08-05 17:07:16 -07:00
  • 8edbcf5209 Improve performance on some lowend GPUs. comfyanonymous 2024-08-05 16:24:04 -04:00
  • e545a636ba This probably doesn't work anymore. comfyanonymous 2024-08-05 12:31:12 -04:00
  • 33e5203a2a Don't cache index.html (#4211) bymyself 2024-08-05 09:25:28 -07:00
  • a178e25912 Fix Flux FP64 math on XPU (#4210) a-One-Fan 2024-08-05 08:26:20 +03:00
  • 78e133d041 Support simple diffusers Flux loras. comfyanonymous 2024-08-04 21:59:42 -04:00
  • 7afa985fba Correct spelling 'token_weight_pars_t5' to 'token_weight_pairs_t5' (#4200) Silver 2024-08-04 23:10:02 +02:00
  • ddb6a9f47c Set the step in EmptySD3LatentImage to 16. comfyanonymous 2024-08-04 15:59:02 -04:00
  • 3b71f84b50 ONNX tracing fixes. comfyanonymous 2024-08-04 15:45:43 -04:00
  • 0a6b008117 Fix issue with some custom nodes. comfyanonymous 2024-08-04 10:03:33 -04:00
  • 56f3c660bf ModelSamplingFlux now takes a resolution and adjusts the shift with it. comfyanonymous 2024-08-04 04:06:00 -04:00
  • f7a5107784 Fix crash. comfyanonymous 2024-08-03 16:55:38 -04:00
  • 91be9c2867 Tweak lowvram memory formula. comfyanonymous 2024-08-03 16:34:27 -04:00
  • 03c5018c98 Lower lowvram memory to 1/3 of free memory. comfyanonymous 2024-08-03 15:14:07 -04:00
  • 2ba5cc8b86 Fix some issues. comfyanonymous 2024-08-03 15:06:40 -04:00
  • 1e68002b87 Cap lowvram to half of free memory. comfyanonymous 2024-08-03 14:50:20 -04:00
  • ba9095e5bd Automatically use fp8 for diffusion model weights if: comfyanonymous 2024-08-03 13:45:19 -04:00
  • f123328b82 Load T5 in fp8 if it's in fp8 in the Flux checkpoint. comfyanonymous 2024-08-03 12:39:33 -04:00
  • 63a7e8edba More aggressive batch splitting. comfyanonymous 2024-08-03 11:53:30 -04:00
  • 0eea47d580 Add ModelSamplingFlux to experiment with the shift value. comfyanonymous 2024-08-03 03:54:38 -04:00
  • 7cd0cdfce6 Add advanced model merge node for Flux model. comfyanonymous 2024-08-02 23:20:30 -04:00
  • ea03c9dcd2 Better per model memory usage estimations. comfyanonymous 2024-08-02 18:08:21 -04:00
  • 3a9ee995cf Tweak regular SD memory formula. comfyanonymous 2024-08-02 17:34:30 -04:00
  • 47da42d928 Better Flux vram estimation. comfyanonymous 2024-08-02 17:02:35 -04:00
  • 17bbd83176 Fix bug loading flac workflow when it contains = character. comfyanonymous 2024-08-02 13:14:28 -04:00
  • bfb52de866 Lower SAG scale step for finer control (#4158) fgdfgfthgr-fox 2024-08-03 02:29:03 +12:00
  • eca962c6da Add FluxGuidance node. comfyanonymous 2024-08-02 10:24:53 -04:00
  • c1696cd1b5 Add missing import (#4174) Jairo Correa 2024-08-02 10:34:12 -03:00
  • 369f459b20 Fix no longer working on old pytorch. comfyanonymous 2024-08-01 22:19:53 -04:00
  • ce9ac2fe05 Fix clip_g/clip_l mixup (#4168) Alexander Brown 2024-08-01 18:40:56 -07:00
  • e638f2858a Hack to make all resolutions work on Flux models. comfyanonymous 2024-08-01 21:03:26 -04:00
  • a531001cc7 Add CLIPTextEncodeFlux. comfyanonymous 2024-08-01 18:53:25 -04:00
  • d420bc792a Tweak the memory usage formulas for Flux and SD. comfyanonymous 2024-08-01 17:49:46 -04:00
  • d965474aaa Make ComfyUI split batches a higher priority than weight offload. comfyanonymous 2024-08-01 16:39:59 -04:00
  • 1c61361fd2 Fast preview support for Flux. comfyanonymous 2024-08-01 16:28:11 -04:00
  • a6decf1e62 Fix bfloat16 potentially not being enabled on mps. comfyanonymous 2024-08-01 16:18:14 -04:00
  • 48eb1399c0 Try to fix mac issue. comfyanonymous 2024-08-01 13:41:27 -04:00
  • b4f6ebb2e8 Rename UNETLoader node to "Load Diffusion Model". comfyanonymous 2024-08-01 13:33:30 -04:00
  • d7430a1651 Add a way to load the diffusion model in fp8 with UNETLoader node. comfyanonymous 2024-08-01 13:28:41 -04:00
  • f2b80f95d2 Better Mac support on flux model. comfyanonymous 2024-08-01 12:55:28 -04:00
  • 1aa9cf3292 Make lowvram more aggressive on low memory machines. comfyanonymous 2024-08-01 12:11:57 -04:00
  • 2f88d19ef3 Add link to Flux examples to readme. comfyanonymous 2024-08-01 11:48:19 -04:00
  • eb96c3bd82 Fix .sft file loading (they are safetensors files). comfyanonymous 2024-08-01 11:32:58 -04:00
  • 5f98de7697 Load flux t5 in fp8 if weights are in fp8. comfyanonymous 2024-08-01 11:05:56 -04:00
  • 8d34211a7a Fix old python versions no longer working. comfyanonymous 2024-08-01 09:57:01 -04:00
  • 1589b58d3e Basic Flux Schnell and Flux Dev model implementation. comfyanonymous 2024-08-01 04:03:59 -04:00
  • 7ad574bffd Mac supports bf16 just make sure you are using the latest pytorch. comfyanonymous 2024-08-01 09:42:17 -04:00
  • e2382b6adb Make lowvram less aggressive when there are large amounts of free memory. comfyanonymous 2024-08-01 03:58:58 -04:00
  • c24f897352 Fix to get fp8 working on T5 base. comfyanonymous 2024-07-31 02:00:19 -04:00
  • a5991a7aa6 Fix hunyuan dit text encoder weights always being in fp32. comfyanonymous 2024-07-31 01:34:57 -04:00
  • 2c038ccef0 Lower CLIP memory usage by a bit. comfyanonymous 2024-07-31 01:32:35 -04:00
  • b85216a3c0 Lower T5 memory usage by a few hundred MB. comfyanonymous 2024-07-31 00:52:34 -04:00
  • 82cae45d44 Fix potential issue with non clip text embeddings. comfyanonymous 2024-07-30 14:20:28 -04:00
  • 25853d0be8 Use common function for casting weights to input. comfyanonymous 2024-07-30 05:03:20 -04:00
  • 79040635da Remove unnecessary code. comfyanonymous 2024-07-30 05:01:34 -04:00
  • 66d35c07ce Improve artifacts on hydit, auraflow and SD3 on specific resolutions. comfyanonymous 2024-07-29 20:27:40 -04:00
  • c75b50607b Less confusing exception if pillow() function fails. comfyanonymous 2024-07-29 11:15:37 -04:00
  • 4ba7fa0244 Refactor: Move sd2_clip.py to text_encoders folder. comfyanonymous 2024-07-28 01:19:20 -04:00
  • ab76abc767 Active workflow use primary fg color (#4090) bymyself 2024-07-27 20:34:19 -07:00
  • 9300058026 Add dpmpp_2s_ancestral as custom sampler (#4101) Silver 2024-07-27 22:19:50 +02:00
  • f82d09c9b4 Update packaging workflow. comfyanonymous 2024-07-27 04:48:19 -04:00
  • e6829e7ac5 Add a way to set custom dependencies in the release workflow. comfyanonymous 2024-07-27 04:41:46 -04:00
  • 07f6a1a685 Handle case in the updater when master branch is not in local repo. comfyanonymous 2024-07-27 03:15:22 -04:00
  • e746965c50 Update nightly package workflow. comfyanonymous 2024-07-27 01:20:18 -04:00
  • 45a2842d7f Set stable releases as a prerelease initially. comfyanonymous 2024-07-26 14:52:00 -04:00
  • 17b41f622e Change windows standalone URL to stable release. (#4065) Robin Huang 2024-07-26 11:37:40 -07:00
  • cf4418b806 Don't treat Bert model like CLIP. comfyanonymous 2024-07-26 13:07:39 -04:00
  • 6225a7827c Add CLIPTextEncodeHunyuanDiT. comfyanonymous 2024-07-26 13:04:48 -04:00
  • b6779d8df3 Fix undo incorrectly undoing text input (#4114) filtered 2024-07-27 02:25:42 +10:00
  • 8328a2d8cd Let hunyuan dit work with all prompt lengths. comfyanonymous 2024-07-26 12:11:32 -04:00
  • afe732bef9 Hunyuan dit can now accept longer prompts. comfyanonymous 2024-07-26 11:52:58 -04:00
  • a9ac56fc0d Own BertModel implementation that works with lowvram. comfyanonymous 2024-07-26 04:32:33 -04:00
  • 25b51b1a8b Hunyuan DiT lora support. comfyanonymous 2024-07-25 22:42:54 -04:00
  • 61a2b00bc2 Add HunyuanDiT support to readme. comfyanonymous 2024-07-25 19:06:43 -04:00
  • a5f4292f9f Basic hunyuan dit implementation. (#4102) comfyanonymous 2024-07-25 18:21:08 -04:00
  • f87810cd3e Let tokenizers return weights to be stored in the saved checkpoint. comfyanonymous 2024-07-25 10:52:09 -04:00
  • 10c919f4c7 Make it possible to load tokenizer data from checkpoints. comfyanonymous 2024-07-24 16:43:53 -04:00
  • ce80e69fb8 Avoid loading the dll when it's not necessary. comfyanonymous 2024-07-24 13:50:34 -04:00
  • 19944ad252 Add code to fix issues with new pytorch version on the standalone. comfyanonymous 2024-07-24 12:49:29 -04:00
  • 10b43ceea5 Remove duplicate code. comfyanonymous 2024-07-24 01:12:59 -04:00
  • 0a4c49c57c Support MT5. comfyanonymous 2024-07-23 15:35:28 -04:00
  • 88ed893034 Allow SPieceTokenizer to load model from a byte string. comfyanonymous 2024-07-23 14:17:42 -04:00
  • 334ba48cea More generic unet prefix detection code. comfyanonymous 2024-07-23 14:13:32 -04:00