Commit Graph

Select branches

Hide Pull Requests

auto_convert

gemma_tenc

main

stable

Mono Color

• bc5223b0e3 Fix qtype logging on newer gguf-py City 2025-03-16 20:14:33 +01:00
• 4aebf1bf2c Add compat mode pig City 2025-03-16 19:38:40 +01:00
• 5aa6e98ad7 Merge pull request #224 from city96/tokenizer_convert City 2025-03-14 17:53:02 +01:00
• ad69decc8d Update requirements.txt City 2025-03-14 17:43:05 +01:00
• 31dffa061b Add error message for missing packages City 2025-03-14 17:37:14 +01:00
• 296b0c1419 Clean up field logic City 2025-03-14 17:11:47 +01:00
• d99b2d931f force dequant on umT5 to prevent OOM City 2025-03-14 16:56:48 +01:00
• d40e9bdcfe Use args for 5d fix City 2025-03-14 16:48:01 +01:00
• 9d0a0e317e ltxv bf16 fix City 2025-03-07 22:37:17 +01:00
• c3a6103af5 hyvid subkey load fix City 2025-03-07 22:36:51 +01:00
• 01f6e3ff63 spiece tokenizer from gguf metadata City 2025-03-05 04:44:53 +01:00
• 929a9e5f0d Keep img_emb in F32 for wan i2v City 2025-02-27 05:58:01 +01:00
• 32a8eab8f9 Merge branch 'main' into convert_refactor_new City 2025-02-27 03:19:19 +01:00
• 298192ed60 Wan model arch City 2025-02-26 07:40:44 +01:00
• cd361cf0e3 Update README.md City 2025-02-26 07:39:01 +01:00
• 8832d37df2 Solution for patch files line endings and update for /tools README.md (#202) YarvixPA 2025-02-26 01:35:42 -05:00
• 355b43228c Wan2.1 convert logic City 2025-02-26 07:06:54 +01:00
• e024aab10d Torch compile possible workaround City 2025-02-23 18:49:50 +01:00
• abdfec54c4 Hunyuan Jank City 2025-02-20 05:10:27 +01:00
• 2d46bcd61b Sanity checks in conversion code City 2025-01-08 17:45:37 +01:00
• 8d313a677a Unused/old logic City 2025-01-08 17:17:49 +01:00
• 5875c52f59 Unpin numpy version City 2025-01-08 16:30:17 +01:00
• 8098c7149f Add PixArt clip type City 2025-01-04 21:46:23 +01:00
• 3dc384b233 More robust folder path logic City 2024-12-26 21:42:33 +01:00
• c6ad4f2363 Account for not having quantized weights City 2024-12-22 19:00:03 +01:00
• 51fa2cbb5c Account for dequantization VRAM to avoid OOM City 2024-12-22 18:09:51 +01:00
• eaffa0a7d3 Change clip workaround for SD3 City 2024-12-22 15:24:14 +01:00
• 8d1b56c5e2 Loader logic for LLaMA models City 2024-12-20 13:39:17 +01:00
• 0a1723baae Split loader into separate file City 2024-12-20 13:03:27 +01:00
• 0b65261cc4 HunyuanVideo City 2024-12-17 07:33:42 +01:00
• 65a7c895bb Better propagation fix City 2024-12-14 00:38:11 +01:00
• 23b9817081 LTX-Video City 2024-12-08 00:48:59 +01:00
• 851564739d Updated Intel Arc fix City 2024-12-05 03:16:23 +01:00
• 39102df73f Stop custom tensor class propagating past dequant City 2024-12-04 16:14:54 +01:00
• 6e4790f903 Unused workaround City 2024-12-04 15:54:20 +01:00
• a8b5f5d892 Less fragile patch logic City 2024-12-04 02:09:01 +01:00
• 4a84328841 Add AuraFlow support City 2024-11-23 06:51:29 +01:00
• 57562d54a5 Add ltxv clip type City 2024-11-23 00:10:27 +01:00
• 51c9713400 Fix sd1 not using shape fix City 2024-11-21 02:24:40 +01:00
• 8e898fad4c Add mochi to CLIPLoaderGGUF options (#154) RandomGitUser321 2024-11-06 08:04:04 -05:00
• 9833348005 Fix for text encoder folder name change City 2024-11-03 14:03:42 +01:00
• 6561064dcf Add SD3.5 links City 2024-10-23 04:38:39 +02:00
• b70240508f Fix SD3 patch metadata/tensor issue City 2024-10-23 03:16:14 +02:00
• 4d7bb93397 Add initial SD3 lcpp patch + instruction City 2024-10-23 01:39:45 +02:00
• 0ef158d793 Add SD3 to loader arch list City 2024-10-22 22:00:37 +02:00
• 68ad5fb214 Update README.md City 2024-10-15 04:11:45 +02:00
• d090a910ac Rewrite convert.py to work properly with dev de-distill City 2024-10-15 03:56:54 +02:00
• d2aaeb0f13 Always init tenc on offload device City 2024-09-17 15:26:39 +02:00
• 454955ead3 Skip initialization for nn.Linear City 2024-09-09 01:37:11 +02:00
• ac96699b7a Update README.md City 2024-09-06 18:41:49 +02:00
• 0342f0a183 never double patch City 2024-09-06 18:10:02 +02:00
• b46c206dd7 double patch issue fix for lowvram City 2024-09-05 05:32:01 +02:00
• 717a0e1d0b Temporarily disable instant patch logic City 2024-09-03 17:00:20 +02:00
• c8923a4cd7 mmap unload fix for lowvram City 2024-09-03 16:37:35 +02:00
• 6dbb4ba989 Temp fix for disable smart memory City 2024-09-03 05:28:49 +02:00
• 69f0daffc5 Merge pull request #92 from city96/ops_refactor City 2024-09-03 03:13:54 +02:00
• 929d15454d Rename ggml_forward City 2024-09-02 03:58:26 +02:00
• 35007f2659 Refactor model loading City 2024-09-02 03:44:07 +02:00
• df3378da6b Detect quant City 2024-09-02 03:17:16 +02:00
• cd343d1426 Merge pull request #80 from blepping/feat_load_stable_diffusion.cpp City 2024-09-02 02:51:40 +02:00
• fa1671d63e Sync with upstream blepping 2024-09-01 18:44:31 -06:00
• 379175e7bf Refactor tensor type checking a bit more blepping 2024-09-01 18:30:47 -06:00
• ba8ce44b4c Rename move_patch function City 2024-09-02 02:22:32 +02:00
• e19fe8e898 Rename ggml_check_delegate to is_ggml_quantized, refactor accordingly blepping 2024-09-01 18:04:31 -06:00
• c7b902bf9a Patch load/unload fix City 2024-09-02 02:00:10 +02:00
• 43427eff52 Better patch clear logic City 2024-09-02 01:37:16 +02:00
• 3bad1b4e04 Fix loading GGUF t5xxl blepping 2024-09-01 06:46:50 -06:00
• a0cb97eceb Add LayerNorm and GroupNorm ops blepping 2024-09-01 05:25:32 -06:00
• 436ecb4913 Merge upstream ops_refactor blepping 2024-09-01 03:02:59 -06:00
• ed0702fefc Merge branch 'city96:main' into feat_load_stable_diffusion.cpp blepping 2024-09-01 02:47:16 -06:00
• 7f3ced6c3e Consider FP8 as source City 2024-09-01 00:56:30 +02:00
• 6618801f1c Don't dequantize token embed on load City 2024-08-30 06:25:19 +02:00
• c231239fdf cleanup City 2024-08-30 03:47:52 +02:00
• ec24e75d89 Remove unused check City 2024-08-30 03:40:15 +02:00
• f40b149bc5 Add BF16 support City 2024-08-30 03:39:57 +02:00
• 826bd1804c Make ops less jank City 2024-08-30 03:25:57 +02:00
• b88cb9a344 Remove require_prefix in gguf_sd_loader blepping 2024-08-28 12:50:39 -06:00
• e814fc1264 Revert ggml_tensor_name addition blepping 2024-08-28 12:34:08 -06:00
• abe5c6a6c9 Fix weights not being copied for lowvram patch City 2024-08-28 20:23:08 +02:00
• f7d8fbe740 Workaround for stable-diffusion.cpp SDXL GGUF files blepping 2024-08-28 11:56:51 -06:00
• 6ae0c733b0 Support loading u-net from stable-diffusion.cpp GGUF files blepping 2024-08-28 10:34:58 -06:00
• 3d9869240b Always patch weight City 2024-08-28 17:12:24 +02:00
• 258742c369 Fix Conv2D with bias City 2024-08-26 17:49:18 +02:00
• a9f6359a23 Merge pull request #63 from blepping/feat_sd1_sdxl City 2024-08-25 23:46:43 +02:00
• df6f1ab39c Sync with upstream blepping 2024-08-25 15:32:45 -06:00
• bfecd32f8c Rewrite llama.cpp patch City 2024-08-25 22:32:55 +02:00
• c1693d0f55 Use a GGUF array for orig shape metadata blepping 2024-08-25 13:18:35 -06:00
• 0aec56ec4c Split orig shape metadata into individual fields blepping 2024-08-25 12:00:28 -06:00
• 7413de86e6 Revert tensor construction change in dequantize_tensor blepping 2024-08-25 10:35:39 -06:00
• f3cf802dd6 Cleanups, refactoring blepping 2024-08-25 07:53:19 -06:00
• 2832969aea Sync with main blepping 2024-08-25 04:35:52 -06:00
• 0ba6f64974 Advanced loader with optimizations City 2024-08-24 18:39:34 +02:00
• 90a63c0c0e Fix for Intel Arc City 2024-08-24 08:17:06 +02:00
• c93e622a19 SD1/SDXL fixes blepping 2024-08-23 20:19:48 -06:00
• 489aaecf7a Show the correct type when using fallback quantization blepping 2024-08-23 12:37:31 -06:00
• cea1b47ab3 Add support for SD1x and SDXL models blepping 2024-08-23 11:59:21 -06:00
• 377862a06d Cast to device City 2024-08-23 03:23:49 +02:00
• 3aafaf9133 Fix LoRA loading on new versions City 2024-08-23 03:20:15 +02:00
• 6b499ec49e Hotfix to work as normal tensor City 2024-08-23 02:13:48 +02:00
• 1743f8f1fa Update README.md City 2024-08-22 21:37:13 +02:00