Commit Graph
963 Commits
Author SHA1 Message Date
leejet 46362fc4c0 feat: add automatic direct convolution selection (#2134) master-963-46362fc 2026-10-13 23:25:19 +08:00
leejet d763dae3b8 feat: add per-module automatic flash attention selection (#2132) master-962-d763dae 2026-10-13 22:58:45 +08:00
losewayy 60423ab5f6 fix: reclaim before accepting wrapped Vulkan iGPU memory estimates (#2124) master-961-60423ab 2026-10-13 21:10:05 +08:00
Robert Oswald 7e3e20a50b ci: add ROCm Dockerfile (#2094) master-960-7e3e20a 2026-10-13 21:09:13 +08:00
leejet d637cfb935 feat: add Qwen Image 2.1 Turbo sigma preset (#2130) master-959-d637cfb 2026-10-12 23:09:42 +08:00
Gabriele Lozupone 792bcd344c feat: add MoSR upscaler support (#2128) master-958-792bcd3 2026-10-12 22:38:38 +08:00
leejet 2988060a43 feat: add TAEQI2.1 support for Qwen Image 2.1 (#2123) 2026-10-11 00:01:18 +08:00
Daniele 1b0ba10893 feat: run conditional and unconditional CFG in one batched UNet forward (#2085) master-956-1b0ba10 2026-10-10 23:35:58 +08:00
losewayy b7a74e68d7 fix: demote resident params to disk residency when memory reclamation fails (#2119) master-955-b7a74e6 2026-10-10 23:15:16 +08:00
Mario Limonciello 05845fde41 ci: fix ROCm ccache build caching and update to ROCm 10.1.0 (#1890) master-954-05845fd 2026-10-10 22:49:28 +08:00
leejet 0e1d733b7a docs: restrict README UI listings to open-source projects 2026-10-10 22:40:06 +08:00
losewayy ce95c09ce1 fix: halve VAE decode tile size on OOM retry instead of clamping to 256px (#2120) 2026-10-10 22:33:37 +08:00
leejet f89d9b13d7 feat: add Iris-3B text-to-image support (#2117) master-951-f89d9b1 2026-10-10 01:39:49 +08:00
leejet f2f168fa70 fix: preserve singleton output dimension when merging LoRA (#2116) master-950-f2f168f 2026-10-10 01:12:03 +08:00
leejet 7867f6da4a fix: bump RPC protocol patch version for rope apply op (#2115) master-949-7867f6d 2026-10-10 00:21:25 +08:00
leejet 228c707fde perf: use ggml rope apply op on supported backends (#2113) master-948-228c707 2026-10-09 04:32:36 +08:00
Daniel K. Gralak e16d26a96a fix: align SD3 encoder token chunks before conditioning (#2111) master-947-e16d26a 2026-10-08 23:52:03 +08:00
mikemikimike a4a966990d fix: avoid scanning cwd during server startup (#2086) master-946-a4a9669 2026-10-08 23:40:43 +08:00
leejet a1ded76da5 chore: resolve MSVC warnings in ggml extensions master-945-a1ded76 2026-10-06 20:10:36 +08:00
leejet 8cc5ad8f1c sync: update frontend master-944-8cc5ad8 2026-10-06 20:04:34 +08:00
leejet 4a892293ac docs: document video reference image support and constraints 2026-10-06 20:03:52 +08:00
leejet fd43b542a6 refactor: escape newlines in prompt logs and clarify source separators (#2106) master-942-fd43b54 2026-10-06 19:51:12 +08:00
losewayy 0963357214 fix: keep MiniMax-H3 VAE weights resident across temporal chunks (#2103) master-941-0963357 2026-10-06 19:40:48 +08:00
stduhpfandleejet 1da55ebde8 feat: enable sdapi latent2RGB previews (#2096)
Co-authored-by: leejet <leejet714@gmail.com>
master-940-1da55eb
2026-10-06 19:28:30 +08:00
leejet fb8ee71f55 refactor: unify circular and non-circular tiling (#2105) master-939-fb8ee71 2026-10-06 19:20:44 +08:00
stduhpfandleejet 96f67fe631 feat: Server: Add support for previews via sdcpp API (#2093)
Co-authored-by: leejet <leejet714@gmail.com>
master-938-96f67fe
2026-10-06 19:20:02 +08:00
leejet 203a882a28 refactor: shorten log levels and move source locations to the end (#2104) master-937-203a882 2026-10-06 18:26:49 +08:00
mcxu 019bcfc0ae fix: correct non-circular tile placement and blending (#2088) master-936-019bcfc 2026-10-06 18:23:41 +08:00
Wagner Bruna c150a6bafe fix: correct step count adjustment for custom sigmas list (#2084) master-935-c150a6b 2026-10-06 17:55:00 +08:00
Wagner Brunaandleejet f4b5326da7 feat: support for merging LoRA weights on model conversion (#2079)
Co-authored-by: leejet <leejet714@gmail.com>
master-934-f4b5326
2026-10-06 17:50:16 +08:00
Aleksei Pakhomov bbdcad48dd feat: add Z-Image L2P support (#2075) master-933-bbdcad4 2026-10-06 17:23:21 +08:00
losewayyandleejet b6947b8bdc fix: avoid f16 overflow in Z-Image quantized matmuls on CUDA (#2074)
Co-authored-by: leejet <leejet714@gmail.com>
master-932-b6947b8
2026-10-06 17:19:56 +08:00
Daniele 9db0db6570 perf: avoid redundant permute+cont in non-interleaved RoPE (#2102) master-931-9db0db6 2026-10-06 16:54:55 +08:00
Harkirat Gill d4bfcd5d78 ci: bundle HIP runtime libraries with Windows ROCm release (#2090) 2026-10-06 16:42:02 +08:00
leejet 3f8527a46c feat: enable hip INT8 tensorwise matmul and convrot (#2071) master-929-3f8527a 2026-09-27 23:24:07 +08:00
leejet 42ab1c12f1 fix: keep scaled INT8 convrot matmuls on Vulkan (#2070) master-928-42ab1c1 2026-09-27 20:43:06 +08:00
leejet 27f7c430c0 fix: update ggml to fix INT8 convrot backend fallback (#2069) master-927-27f7c43 2026-09-27 20:42:31 +08:00
leejet ede32a621e fix: load INT8 convrot LLM embeddings correctly (#2067) master-926-ede32a6 2026-09-27 19:46:21 +08:00
leejet 47e83d7136 sync: update ggml master-925-47e83d7 2026-09-27 18:40:09 +08:00
leejet f8890b9db3 feat: add Ming-Image Design support (#2063) master-924-f8890b9 2026-09-27 17:24:13 +08:00
akleine 61a83b85cc docs: update hugging face links for MageFlow diffusion and VAE models (#2065) 2026-09-27 17:22:03 +08:00
yi111 9947eebec8 fix: handle filesystem errors during lora and upscaler cache scan (#2064) master-922-9947eeb 2026-09-27 17:21:19 +08:00
losewayyandleejet 168f7b88b9 fix: improve GGUF metadata parsing and reader selection (#2062)
Co-authored-by: leejet <leejet714@gmail.com>
master-921-168f7b8
2026-09-27 16:24:07 +08:00
leejet 2f886889e6 refactor: align PixArt weights with upstream layout (#2061) master-920-2f88688 2026-09-26 02:03:05 +08:00
leejet 19bbbca1c7 refactor: define VAE tile dimensions in image pixels (#2059) master-919-19bbbca 2026-09-25 18:20:32 +08:00
losewayy 39ada0863b feat: add PixArt model family support (#2047) master-918-39ada08 2026-09-25 18:18:55 +08:00
leejet 4c3cf7543d fix: load safetensors index shards without recursion (#2058) master-917-4c3cf75 2026-09-25 16:40:20 +08:00
leejet 510bccf330 fix: map Qwen Image 2.1 LoRAs to fused MLP weights (#2057) 2026-09-25 16:30:34 +08:00
yi111 5e7b29131e fix: keep the server frontend install out of a parent pnpm workspace (#2056) master-915-5e7b291 2026-09-25 16:11:21 +08:00
leejet 0a9340c599 fix: scale Qwen Image 2.1 VAE convolutions (#2054) master-914-0a9340c 2026-09-25 15:52:35 +08:00