github/ai-toolkit
1
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity

Commit Graph

  • 691ddf434e Add Qwen3.6 VL captioner. Jaret Burkett 2026-07-15 07:02:33 -06:00
  • 18da85153b Disable xet by default. Seems to be causing a lot of hanging issues. Jaret Burkett 2026-07-14 10:42:03 -06:00
  • cf0db39ede Fix some errors for specific quants. Enable saving quantizations. Jaret Burkett 2026-07-14 07:25:09 -06:00
  • abba6b5845 Show better errors on captioner Jaret Burkett 2026-07-14 07:19:03 -06:00
  • 8b5bf25b13 Add convrot quants to captioner Jaret Burkett 2026-07-14 07:03:10 -06:00
  • 676b4f3c4c Add Automagic3 optimizer to the ui Jaret Burkett 2026-07-14 06:39:10 -06:00
  • 0d53e5e1f9 Fix compile with lokr on convrot Jaret Burkett 2026-07-13 19:24:57 -06:00
  • a5f857ddb0 Added patch from Fatalis to fix lokr offloading with convrot Jaret Burkett 2026-07-13 18:48:56 -06:00
  • 28f2c0acbe Move z_image over to the new modeling class Jaret Burkett 2026-07-13 17:12:30 -06:00
  • dcb3b329b2 Fix issue with the graph with sparse data when shrinking smoothing Jaret Burkett 2026-07-13 17:10:38 -06:00
  • 1f7d608e20 Show sparse datapoints on the loss graph. Jaret Burkett 2026-07-13 10:59:57 -06:00
  • 7602e476eb Exclude sensative layers from quantization in krea Jaret Burkett 2026-07-13 10:14:53 -06:00
  • 28b05ee4ed Added convrotbitnet 1.58 bit quantization Jaret Burkett 2026-07-13 07:52:48 -06:00
  • a259fa07cd Make convrot ui clearer Jaret Burkett 2026-07-13 06:44:27 -06:00
  • 0b62e516cc Version Bump Jaret Burkett 2026-07-13 06:34:21 -06:00
  • b6ff367633 Convrot improvements. Add argitrary bit packed byte quantizations. Jaret Burkett 2026-07-13 06:33:54 -06:00
  • 64663c8575 Add Boogu to the readme. Jaret Burkett 2026-07-11 21:03:00 -06:00
  • 4625406093 WIP working on convrot offloading Jaret Burkett 2026-07-11 15:37:28 -06:00
  • 1d1e21177a Add ostris linear layer to linear layer searches. Jaret Burkett 2026-07-11 14:21:16 -06:00
  • 095d6e7418 Add convrot4 and convrot8 quantizations to the ui. A lot of improvements to the convrot quantization methods. Jaret Burkett 2026-07-11 13:49:41 -06:00
  • 933ca1c517 Apply w8a8 on the forward when training with convrot for more speed. Fix issue where quantizing a model had a pin memory leak. Jaret Burkett 2026-07-11 10:53:50 -06:00
  • 065ac27353 Fix issue with compiling convrot on some models Jaret Burkett 2026-07-11 09:26:12 -06:00
  • 96a3a06111 Added initial experimental support for convrot4 and convrot8 quantization. Jaret Burkett 2026-07-10 11:01:14 -06:00
  • 6fac83d068 Fix casting issue with orbit quant Jaret Burkett 2026-07-10 10:23:38 -06:00
  • 71c75357eb Add cached conditioning recovery to wan 22 5b model Jaret Burkett 2026-07-10 10:22:24 -06:00
  • ad07b06de5 Use cached first frame for wan22_5 model Jaret Burkett 2026-07-10 10:21:34 -06:00
  • 886c2aec57 Allow for vae tiling onle without low vram on wan models with a model kwarg Jaret Burkett 2026-07-10 09:42:52 -06:00
  • fe82487187 Add tiling on vae decode for qwen image models when low vram flag is on Jaret Burkett 2026-07-10 09:07:29 -06:00
  • e7951ad29e Exclude sensative layers from quantization on wan models Jaret Burkett 2026-07-10 08:56:05 -06:00
  • 883d60eb71 Do vae tiling when decoding wan models with low vram active. Jaret Burkett 2026-07-10 08:02:32 -06:00
  • fed9357234 Fix ui to show kv cache for krea2 edit raw Jaret Burkett 2026-07-09 15:48:46 -06:00
  • 5a9b5bde3f Added experimental orbit quant Jaret Burkett 2026-07-09 15:14:51 -06:00
  • a4bbe167ce Added reference token attention isolation (kv_cache) for Krea2 edit training. Same training cost with significant inference speed up. 2x inference speedup. Jaret Burkett 2026-07-09 12:00:55 -06:00
  • 6233efe1bb Fix sampling issue on zimage turbo and krea2 turbo models Jaret Burkett 2026-07-08 12:34:39 -06:00
  • dd08579eda Add ability to pull control images from same folder group Jaret Burkett 2026-07-08 05:21:31 -06:00
  • 7bceec3b07 Apply a loss mask for first frame conditioning for wan and ltx Jaret Burkett 2026-07-07 09:05:53 -06:00
  • bd93a312bc Load more types of comfy ui style quants for ltx2 Jaret Burkett 2026-07-07 09:04:51 -06:00
  • 17bc302d13 Make log calls blocking to prevent duplicates. Jaret Burkett 2026-07-06 07:38:20 -06:00
  • 6c0d1c4679 Improve latency of the job list endpoint Jaret Burkett 2026-07-04 13:22:09 -06:00
  • 3a94591c89 Disable text encoder unload with krea2 o-edit Jaret Burkett 2026-07-04 09:17:15 -06:00
  • b1e1a834d4 Added support to train Krea2 as an edit model Jaret Burkett 2026-07-04 09:12:39 -06:00
  • f63221e577 Replace all sync functions with async to allow more parallel api calls Jaret Burkett 2026-07-01 17:12:43 -06:00
  • 48781f900b Improve the loading and transfer speed of the dataset file lists Jaret Burkett 2026-07-01 16:40:01 -06:00
  • b36a8e9c4b Only pull the new log bytes from the ui to improve performance Jaret Burkett 2026-07-01 14:11:23 -06:00
  • 733e14cb58 Fix issue where the window would scroll to and hilight custom items in the select box Jaret Burkett 2026-07-01 13:57:56 -06:00
  • 4e50535478 Allow models to provide an additional loss Jaret Burkett 2026-06-28 09:49:36 -06:00
  • 1e12b6b73f Remove duplicate save when triggering save next step Jaret Burkett 2026-06-28 07:18:16 -06:00
  • 7ee1f98f6d Allow loading and saving z_image in comfy safetensors format. Jaret Burkett 2026-06-28 07:15:49 -06:00
  • c97fc9973a Fix loading pretraied lora when merging network on save Jaret Burkett 2026-06-28 06:04:36 -06:00
  • f8667f0334 Save Krea2 even if quantized Jaret Burkett 2026-06-28 06:03:37 -06:00
  • df6ea4263d Version bump Jaret Burkett 2026-06-26 13:02:10 -06:00
  • ad87aacec0 Add ability to set certain layers to full for loras Jaret Burkett 2026-06-26 13:01:48 -06:00
  • 4a99ddabad Fix breaking change with diffusers qwen image Jaret Burkett 2026-06-26 10:04:35 -06:00
  • 5f04ae7ad5 Rework merge_network_on_save to handle dequantization on merging and saving for much more efficient full finetuning. Jaret Burkett 2026-06-25 13:19:38 -06:00
  • 6ecff36f26 Add a way to do full rank lora modules of non lora layers such as embeddings, norm, etc Jaret Burkett 2026-06-25 11:55:07 -06:00
  • 4eb0707639 Add a control generation script. Jaret Burkett 2026-06-25 10:14:24 -06:00
  • d14f6e567a Allow individual models to scale the loss after it is calculated. Jaret Burkett 2026-06-25 10:13:58 -06:00
  • f743ccf7ef Save before last sample Jaret Burkett 2026-06-24 09:13:01 -06:00
  • 089e41dd1c Compile improvements - auto cache size, fix fullgraph setting, fix triton detection (#899) PlagueKind 2026-06-24 15:59:41 +02:00
  • d586125b40 Force HF downloads to show progress bars when output is not a TTY (#909) fatalis 2026-06-24 08:55:38 -05:00
  • 7a089fd0d7 Add support for training directly on Krea2 Turbo with a training adapter Jaret Burkett 2026-06-23 20:32:02 -06:00
  • a803611ec1 Enable tiling on vae when decoding with low_vram flag on krea2 Jaret Burkett 2026-06-23 18:38:52 -06:00
  • 724e67d634 Set krea 2 to use new lokr format Jaret Burkett 2026-06-23 13:31:50 -06:00
  • e20b42e84a Add offloading support for krea2 Jaret Burkett 2026-06-23 10:43:36 -06:00
  • 99be3d96a2 Add support for Krea2 (#906) Jaret Burkett (Ostris) 2026-06-23 09:18:14 -06:00
  • af594061ab Add ability to do hidden states with tipsv2 Jaret Burkett 2026-06-22 11:00:04 -06:00
  • 820d534d6e Add features for models that may need a non masked loss such as inpainiting. Jaret Burkett 2026-06-22 10:59:17 -06:00
  • c133c55cf5 Fix issue with mask generator revision Jaret Burkett 2026-06-22 08:19:27 -06:00
  • d51463ca52 Add ability to recover from a truncated image file in the dataset Jaret Burkett 2026-06-22 05:07:30 -06:00
  • ba0b3dbb65 Force batch size when bucket is too small by duplicating items in the batch Jaret Burkett 2026-06-21 20:01:29 -06:00
  • dba092fc15 Keep a real git repo in the docker image Jaret Burkett 2026-06-21 06:54:19 -06:00
  • 548a286992 Add captioning monitoring information to the dataset page so you can just stay on that page with all the info there. Jaret Burkett 2026-06-19 08:00:40 -06:00
  • 99f8fd44e3 Add a fallback bbox adjustment when json parsing fails on ideogram4 captioner Jaret Burkett 2026-06-19 07:25:54 -06:00
  • 4af4fb9d58 Add auto_frame_count support for all remaining video models (#897) fatalis 2026-06-19 08:05:18 -05:00
  • 022d1c29e0 Remove color pallet for individual objects in the caption prompts Jaret Burkett 2026-06-19 06:56:18 -06:00
  • 60c1ac6a50 Add support for Boogu Image and Boogu Image Edit Jaret Burkett 2026-06-18 15:05:49 -06:00
  • e886745051 Added additional information on addine new models and some additional gotchas Jaret Burkett 2026-06-18 15:04:50 -06:00
  • 515b0ea5cd Remove transformer log supression from captioner Jaret Burkett 2026-06-18 10:00:30 -06:00
  • e8c828089a Fix issue where gpu sometimes doesnt show on caption modal Jaret Burkett 2026-06-18 09:37:59 -06:00
  • ad49d4ef25 Update example to cover some common issues Jaret Burkett 2026-06-18 07:54:37 -06:00
  • 66f7c06742 Patch away the qwen3vl conv3d that only has slow bf16 kernels Jaret Burkett 2026-06-17 14:29:20 -06:00
  • 92814f9e6d Apply ideogram dynamic shifting to sampling Jaret Burkett 2026-06-16 18:52:16 -06:00
  • 178eb5fbbe Add unconditional lora support so Ideogram 4 inference will more closely resemble the full pipeline results. I pushed a finetuned unconditional lora to the hub as an adapter. Jaret Burkett 2026-06-16 13:27:43 -06:00
  • f6c0104f25 Handle unconditional conditioning for ideogram 4 more in line with example code. Jaret Burkett 2026-06-16 10:23:08 -06:00
  • 86b19589a0 Update the Ideogram 4 prompt generation/parsing/ui to handle the updated format notes better. Jaret Burkett 2026-06-16 09:44:38 -06:00
  • fcccc0fbd2 Add gradient checkpointing to ideogram4 vae Jaret Burkett 2026-06-15 10:00:40 -06:00
  • c730d64478 Added a flag to keep loading the image when latents are cached. Useful for DFE and other methods that target pixelspace losses. Jaret Burkett 2026-06-15 05:31:48 -06:00
  • faa770fc79 Another docker thing? Jaret Burkett 2026-06-14 12:27:05 -06:00
  • 570c806924 Mode docker build work Jaret Burkett 2026-06-14 12:24:55 -06:00
  • ebbb09230b Add requirements base to docker build Jaret Burkett 2026-06-14 12:21:28 -06:00
  • 5df3fb69e3 Rework Docker image for a minimal build/pull/push footprint Jaret Burkett 2026-06-14 12:19:58 -06:00
  • c0d600b5d6 Allow using flash backend for ideogram Jaret Burkett 2026-06-13 15:17:20 -06:00
  • c8cd78b1a4 Allow nested transformer block names for quantization, lora targeting, quantizing Jaret Burkett 2026-06-13 14:33:16 -06:00
  • 17c9279828 Version bump Jaret Burkett 2026-06-13 09:48:26 -06:00
  • 6c3b82696e Add support for PRX Pixel T2I Jaret Burkett 2026-06-13 09:47:53 -06:00
  • a01c83073a Added an example model with docs so people and agents can add models easier. Jaret Burkett 2026-06-13 08:17:06 -06:00
  • 2f91db8363 Defauly to compiling full graph to false Jaret Burkett 2026-06-13 07:28:51 -06:00
  • e908d85f5e Allow quantized unet offload compile and force fullgraph false (#881) PlagueKind 2026-06-13 15:27:28 +02:00
  • 0165fb2ac6 Update npm packages Jaret Burkett 2026-06-12 17:50:47 -06:00
  • c90c400716 Add Ustris Cloud info to the README Jaret Burkett 2026-06-12 12:53:47 -06:00