stduhpf
06accf2b39
feat: add ltxav latent2rgb projection matrix ( #1502 )
2026-05-17 17:52:05 +08:00
stduhpf and leejet
cde20d5ef0
fix: handle stereo format in sd_audio ( #1489 )
...
Co-authored-by: leejet <leejet714@gmail.com >
2026-05-17 16:55:39 +08:00
stduhpf and leejet
9369ab759f
feat: inpaint improvements ( #1357 )
...
* inpaint: get max pixel max instead of single sample
* inpaint: masked diffusion for inpainting models with inflated mask
* refactor tensor interpolate nearest-like reduction paths and generalize max_pool_2d
---------
Co-authored-by: leejet <leejet714@gmail.com >
2026-04-06 00:44:26 +08:00
stduhpf
ed88e215a2
refactor: simplify f8_e5m2_to_f16 function a little bit ( #1358 )
2026-03-30 00:14:33 +08:00
stduhpf
3d33caaef8
fix: make tiling work better when using circular ( #1299 )
2026-03-08 00:25:07 +08:00
stduhpf
aa0b899397
fix: improve handling of VAE decode failures ( #1222 )
2026-02-09 23:29:41 +08:00
stduhpf
9f56833e14
feat: optimize LoKr at runtime ( #1233 )
2026-02-08 00:08:09 +08:00
stduhpf
3959109281
fix: improve LoCon support with other naming conventions ( #1239 )
2026-02-01 20:00:16 +08:00
stduhpf
b87fe13afd
feat: support new chroma radiance "x0_x32_proto" ( #1209 )
2026-01-19 23:51:26 +08:00
stduhpf
60abda56e0
feat: select vulkan device with env variable ( #629 )
2025-12-21 15:35:38 +08:00
stduhpf and leejet
23fce0bd84
feat: add support for Chroma Radiance x0 ( #1091 )
...
* Add x0 Flux pred (+prepare for others)
* Fix convert models with empty tensors
* patch_32 exp support attempt
* improve support for patch_32
* follow official pipeline
---------
Co-authored-by: leejet <leejet714@gmail.com >
2025-12-20 00:55:57 +08:00
stduhpf
9fa7f415df
feat: add taehv support for Wan/Qwen ( #937 )
2025-12-16 22:57:34 +08:00
stduhpf
d96b4152d6
perf: optimize ggml_ext_chunk ( #1084 )
2025-12-14 01:22:41 +08:00
stduhpf
583a02e29e
feat: add Flux.2 VAE proj matrix for previews ( #1017 )
2025-12-09 22:00:45 +08:00
bcc9c0d0b3
feat: handle ggml compute failures without crashing the program ( #1003 )
...
* Feat: handle compute failures more gracefully
* fix Unreachable code after return
Co-authored-by: idostyle <idostyl3@googlemail.com >
* adjust z_image.hpp
---------
Co-authored-by: idostyle <idostyl3@googlemail.com >
Co-authored-by: leejet <leejet714@gmail.com >
2025-12-04 22:04:27 +08:00
stduhpf
edf2cb3846
fix: fix CosXL not being detected ( #989 )
2025-12-03 22:25:02 +08:00
stduhpf
aa2b8e0ca5
fix: patch 1x1 conv weights at runtime ( #986 )
2025-11-19 23:27:23 +08:00
stduhpf
8ecdf053ac
feat: add image preview support ( #522 )
2025-11-10 00:12:02 +08:00
stduhpf
fb748bb8a4
fix: TAE encoding ( #935 )
2025-11-07 22:58:59 +08:00
stduhpf
c42826b77c
fix: resolve multiple inpainting issues ( #926 )
...
* Fix inpainting masked image being broken by side effect
* Fix unet inpainting concat not being set
* Fix Flex.2 inpaint mode crash (+ use scale factor)
2025-11-02 02:10:32 +08:00
stduhpf
77eb95f8e4
docs: fix taesd direct download link ( #917 )
2025-10-28 23:26:23 +08:00
stduhpf
faabc5ad3c
feat: allow models to run without all text encoder(s) ( #645 )
2025-10-25 22:00:56 +08:00
stduhpf
917f7bfe99
fix: support --flow-shift for flux models with default pred ( #913 )
2025-10-23 21:35:18 +08:00
stduhpf
11f436c483
feat: add support for Flux Controls and Flex.2 ( #692 )
2025-10-11 00:06:57 +08:00
2c9b1e2594
feat: add VAE encoding tiling support and adaptive overlap ( #484 )
...
* implement tiling vae encode support
* Tiling (vae/upscale): adaptative overlap
* Tiling: fix edge case
* Tiling: fix crash when less than 2 tiles per dim
* remove extra dot
* Tiling: fix edge cases for adaptative overlap
* tiling: fix edge case
* set vae tile size via env var
* vae tiling: refactor again, base on smaller buffer for alignment
* Use bigger tiles for encode (to match compute buffer size)
* Fix edge case when tile is bigger than latent
* non-square VAE tiling (#3 )
* refactor tile number calculation
* support non-square tiles
* add env var to change tile overlap
* add safeguards and better error messages for SD_TILE_OVERLAP
* add safeguards and include overlapping factor for SD_TILE_SIZE
* avoid rounding issues when specifying SD_TILE_SIZE as a factor
* lower SD_TILE_OVERLAP limit
* zero-init empty output buffer
* Fix decode latent size
* fix encode
* tile size params instead of env
* Tiled vae parameter validation (#6 )
* avoid crash with invalid tile sizes, use 0 for default
* refactor default tile size, limit overlap factor
* remove explicit parameter for relative tile size
* limit encoding tile to latent size
* unify code style and format code
* update docs
* fix get_tile_sizes in decode_first_stage
---------
Co-authored-by: Wagner Bruna <wbruna@users.noreply.github.com >
Co-authored-by: leejet <leejet714@gmail.com >
2025-09-14 16:00:29 +08:00
stduhpf and leejet
c587a43c99
feat: support incrementing ref image index (omni-kontext) ( #755 )
...
* kontext: support ref images indices
* lora: support x_embedder
* update help message
* Support for negative indices
* support for OmniControl (offsets at index 0)
* c++11 compat
* add --increase-ref-index option
* simplify the logic and fix some issues
* update README.md
* remove unused variable
---------
Co-authored-by: leejet <leejet714@gmail.com >
2025-09-07 22:35:16 +08:00
stduhpf and leejet
141a4b4113
feat: add flow shift parameter (for SD3 and Wan) ( #780 )
...
* Add flow shift parameter (for SD3 and Wan)
* unify code style and fix some issues
---------
Co-authored-by: leejet <leejet714@gmail.com >
2025-09-07 02:16:59 +08:00
stduhpf and leejet
21ce9fe2cf
feat: add support for timestep boundary based automatic expert routing in Wan MoE ( #779 )
...
* Wan MoE: Automatic expert routing based on timestep boundary
* unify code style and fix some issues
---------
Co-authored-by: leejet <leejet714@gmail.com >
2025-09-07 01:44:10 +08:00
stduhpf
4c6475f917
feat: show usage on unknown arg ( #767 )
2025-09-01 21:38:34 +08:00
stduhpf
59080d3ce1
feat: change image dimensions requirement for DiT models ( #742 )
2025-07-28 21:58:17 +08:00
stduhpf and leejet
a772dca27a
feat: add Instruct-Pix2pix/CosXL-Edit support ( #679 )
...
* Instruct-p2p support
* support 2 conditionings cfg
* Do not re-encode the exact same image twice
* fixes for 2-cfg
* Fix pix2pix latent inputs + improve inpainting a bit + fix naming
* prepare for other pix2pix-like models
* Support sdxl ip2p
* fix reference image embeddings
* Support 2-cond cfg properly in cli
* fix typo in help
* Support masks for ip2p models
* unify code style
* delete unused code
* use edit mode
* add img_cond
* format code
---------
Co-authored-by: leejet <leejet714@gmail.com >
2025-07-12 15:36:45 +08:00
stduhpf
dafc32d0dd
feat: add support for f64/i64 and clip_g diffusers model ( #681 )
2025-07-06 23:24:55 +08:00
stduhpf
19fbfd8639
feat: override text encoders for unet models ( #682 )
2025-07-04 22:19:47 +08:00
stduhpf
8d0819c548
fix: actually use embeddings with SDXL ( #657 )
2025-07-03 22:39:57 +08:00
stduhpf
83ef4e44ce
feat: add T5 with llama.cpp naming convention support ( #654 )
2025-07-02 23:13:00 +08:00
stduhpf
9251756086
feat: add CosXL support ( #683 )
2025-07-01 23:13:04 +08:00
stduhpf
ea46fd6948
fix: force zero-initialize output of tiling ( #703 )
2025-07-01 23:01:29 +08:00
b1cc40c35c
feat: add Chroma support ( #696 )
...
---------
Co-authored-by: Green Sky <Green-Sky@users.noreply.github.com >
Co-authored-by: leejet <leejet714@gmail.com >
2025-06-29 23:36:42 +08:00
stduhpf and leejet
c9b5735116
feat: add FLUX.1 Kontext dev support ( #707 )
...
* Kontext support
* add edit mode
---------
Co-authored-by: leejet <leejet714@gmail.com >
2025-06-29 10:08:53 +08:00
stduhpf
3fb275a67b
fix: suport sdxl embedddings ( #621 )
2025-03-09 12:21:23 +08:00
stduhpf
f50a7f66aa
fix: fix race condition causing inconsistent value for decoder_only ( #609 )
2025-03-01 11:49:06 +08:00
stduhpf
85e9a12988
fix: preprocess tensor names in tensor types map ( #607 )
...
Thank you for your contribution
2025-03-01 11:48:04 +08:00
stduhpf
fbd42b6fc1
fix: fix embeddings with quantized models ( #601 )
2025-03-01 11:45:39 +08:00
stduhpf and leejet
69c73789fe
fix: force binary mask for inpaint models ( #589 )
...
Co-authored-by: leejet <leejet714@gmail.com >
2025-02-22 21:29:57 +08:00
stduhpf
f23b803a6b
fix:: unapply current loras properly ( #590 )
2025-02-22 21:22:22 +08:00
stduhpf
1be2491dcf
feat: partial LyCORIS support (tucker decomposition for LoCon + LoHa + LoKr) ( #577 )
2025-02-22 21:19:26 +08:00
stduhpf
e500d95abd
fix: fix rank 1 loras ( #575 )
2025-02-05 16:13:17 +08:00
stduhpf
d9b5942d98
feat: add sdxl v-pred suppport ( #536 )
2025-01-18 13:15:54 +08:00
stduhpf
587a37b2e2
fix: avoid sd2((non inpaint) crash on v-pred check ( #537 )
2025-01-18 13:13:34 +08:00
stduhpf
348a54e34a
feat: use pretty-progress for tensor loading ( #516 )
2024-12-28 13:14:52 +08:00
stduhpf
d50473dc49
feat: support 16 channel tae (taesd/taef1) ( #527 )
2024-12-28 13:13:48 +08:00
stduhpf
0d9d6659a7
fix: fix metal build ( #513 )
2024-12-28 13:06:17 +08:00
stduhpf
8f4ab9add3
feat: support Inpaint models ( #511 )
2024-12-28 13:04:49 +08:00
stduhpf
cc92a6a1b3
feat: support more LoRA models ( #520 )
2024-12-28 12:56:44 +08:00
stduhpf
9148b980be
feat: remove type restrictions ( #489 )
2024-11-30 14:22:15 +08:00
stduhpf and leejet
7ce63e740c
feat: flexible model architecture for dit models (Flux & SD3) ( #490 )
...
* Refactor: wtype per tensor
* Fix default args
* refactor: fix flux
* Refactor photmaker v2 support
* unet: refactor the refactoring
* Refactor: fix controlnet and tae
* refactor: upscaler
* Refactor: fix runtime type override
* upscaler: use fp16 again
* Refactor: Flexible sd3 arch
* Refactor: Flexible Flux arch
* format code
---------
Co-authored-by: leejet <leejet714@gmail.com >
2024-11-30 14:18:53 +08:00
stduhpf
53b415f787
fix: remove default variables in c headers ( #478 )
2024-11-24 18:10:25 +08:00
stduhpf and leejet
6ea812256e
feat: add flux 1 lite 8B (freepik) support ( #474 )
...
* Flux Lite (Freepik) support
* format code
---------
Co-authored-by: leejet <leejet714@gmail.com >
2024-11-23 11:41:30 +08:00
stduhpf
9b1d90bc23
fix: improve clip text_projection support ( #397 )
2024-11-23 11:19:27 +08:00
stduhpf and leejet
65fa646684
feat: add sd3.5 medium and skip layer guidance support ( #451 )
...
* mmdit-x
* add support for sd3.5 medium
* add skip layer guidance support (mmdit only)
* ignore slg if slg_scale is zero (optimization)
* init out_skip once
* slg support for flux (expermiental)
* warn if version doesn't support slg
* refactor slg cli args
* set default slg_scale to 0 (oops)
* format code
---------
Co-authored-by: leejet <leejet714@gmail.com >
2024-11-23 11:15:31 +08:00
stduhpf
14206fd488
fix: fix clip tokenizer ( #383 )
2024-09-02 22:31:46 +08:00
stduhpf
f4c937cb94
fix: add some missing cli args to usage ( #363 )
2024-08-28 00:17:46 +08:00
stduhpf