CCalamitousFelicitousnessfix(lora): apply full-diff norm targets and map z-image's renamed modules
| 文件 | 最后提交记录 | 最后更新时间 |
|---|---|---|
fix(pipelines): restore clip-skip text encoding for transformers 5.6+ The vendored encode_prompt copies in the PAG, APG, ControlNet-XS and differential diffusion pipelines dereference the .text_model wrapper that transformers 5.6 removed from CLIPTextModel, so their clip-skip path crashes on SD1.5 and SDXL TE1. Apply the same getattr(te, 'text_model', te) fix as the core parser. Mirrors upstream diffusers, which still carries this deref in pipeline encode_prompt; only the single-file loader was fixed there. | 2 个月前 | |
strict subpaths Signed-off-by: Vladimir Mandic <mandic00@live.com> | 1 个月前 | |
full codespell coverage Signed-off-by: Vladimir Mandic <mandic00@live.com> | 3 个月前 | |
linting Signed-off-by: Vladimir Mandic <mandic00@live.com> | 2 个月前 | |
strict ruff without exclude Co-authored-by: Copilot <copilot@github.com> Signed-off-by: Vladimir Mandic <mandic00@live.com> | 3 个月前 | |
refactor(model): simplify probe cache handling json_helpers does its own error handling and file locking, so the try/except wrappers, threading locks and isfile checks were dead weight; probe cache file paths are now defined in paths.py. | 1 个月前 | |
linting Signed-off-by: Vladimir Mandic <mandic00@live.com> | 2 个月前 | |
fix(lora): apply te networks before encode and honor lora_apply_te Network activation ran after prompt encoding, so text encoder lora weights never affected embeds on the first generation and the stale result was then served from the embed cache. The trailing unfiltered activate in network_load also overrode the te exclude filter, so the lora_apply_te setting was never honored. - parse and activate networks in process_base before pipeline args are built - activate_filtered gates text encoder components on per-request or global lora_apply_te; used by base, hires, detailer and faceid call sites - network_load accepts activate=False for callers that run their own deactivate/activate sequence with include/exclude - network_activate walks excluded components in restore-only mode so a filtered text encoder reverts to backup instead of keeping stale deltas - loaded_loras cache is single-entry since per-filter entries go stale when the setting toggles - prompt embed cache key includes the effective lora_apply_te value | 1 个月前 | |
better handle onnxruntime Signed-off-by: Vladimir Mandic <mandic00@live.com> | 2 个月前 | |
fix(lora): apply te networks before encode and honor lora_apply_te Network activation ran after prompt encoding, so text encoder lora weights never affected embeds on the first generation and the stale result was then served from the embed cache. The trailing unfiltered activate in network_load also overrode the te exclude filter, so the lora_apply_te setting was never honored. - parse and activate networks in process_base before pipeline args are built - activate_filtered gates text encoder components on per-request or global lora_apply_te; used by base, hires, detailer and faceid call sites - network_load accepts activate=False for callers that run their own deactivate/activate sequence with include/exclude - network_activate walks excluded components in restore-only mode so a filtered text encoder reverts to backup instead of keeping stale deltas - loaded_loras cache is single-entry since per-filter entries go stale when the setting toggles - prompt embed cache key includes the effective lora_apply_te value | 1 个月前 | |
full codespell coverage Signed-off-by: Vladimir Mandic <mandic00@live.com> | 3 个月前 | |
linting Signed-off-by: Vladimir Mandic <mandic00@live.com> | 2 个月前 | |
full codespell coverage Signed-off-by: Vladimir Mandic <mandic00@live.com> | 3 个月前 | |
full codespell coverage Signed-off-by: Vladimir Mandic <mandic00@live.com> | 3 个月前 | |
linting Signed-off-by: Vladimir Mandic <mandic00@live.com> | 2 个月前 | |
linting Signed-off-by: Vladimir Mandic <mandic00@live.com> | 2 个月前 | |
cleanup logger | 6 个月前 | |
fix(lora): apply full-diff norm targets and map z-image's renamed modules A full-weight extraction on Z-Image bound 308 modules and applied 172 of them, silently dropping the rest, and left 71 more unmapped. assign_network_names_to_compvis_modules puts every transformer module in network_layer_mapping but skips stamping network_layer_name on norms, which is the attribute the apply pass keys off. try_load_full bound those modules through the mapping and they then never applied; stamp them loader-locally, as try_load_norm already does. Z-Image also names three module groups differently from the diffusers tree: the qk-norms (q_norm/k_norm vs norm_q/norm_k), and the patch embedder and final layer, which live in ModuleDicts keyed by "{patch_size}-{f_patch_size}" and so carry a key the checkpoint has no notion of. Read that key from the live model rather than hardcoding it. The counts close exactly: 68 qk-norms plus 3 non-block targets are the 71 that went unmapped. | 1 个月前 | |
video model loader Signed-off-by: Vladimir Mandic <mandic00@live.com> | 1 个月前 | |
linting Signed-off-by: Vladimir Mandic <mandic00@live.com> | 2 个月前 | |
fix onnx execution providers Signed-off-by: Vladimir Mandic <mandic00@live.com> | 2 个月前 | |
add experimental openai interface Signed-off-by: Vladimir Mandic <mandic00@live.com> | 2 个月前 | |
fix(pipelines): restore clip-skip text encoding for transformers 5.6+ The vendored encode_prompt copies in the PAG, APG, ControlNet-XS and differential diffusion pipelines dereference the .text_model wrapper that transformers 5.6 removed from CLIPTextModel, so their clip-skip path crashes on SD1.5 and SDXL TE1. Apply the same getattr(te, 'text_model', te) fix as the core parser. Mirrors upstream diffusers, which still carries this deref in pipeline encode_prompt; only the single-file loader was fixed there. | 2 个月前 | |
linting Signed-off-by: Vladimir Mandic <mandic00@live.com> | 2 个月前 | |
RUIF013 updates and formatting | 5 个月前 | |
linting Signed-off-by: Vladimir Mandic <mandic00@live.com> | 2 个月前 | |
update sampler behavior and user definable fallback Signed-off-by: Vladimir Mandic <mandic00@live.com> | 2 个月前 | |
full codespell coverage Signed-off-by: Vladimir Mandic <mandic00@live.com> | 3 个月前 | |
schedulers fix zero-sigma final-step Signed-off-by: Vladimir Mandic <mandic00@live.com> | 2 个月前 | |
set group_size_m to 8 on intel too | 1 个月前 | |
debounce width/height updates Signed-off-by: Vladimir Mandic <mandic00@live.com> | 2 个月前 | |
Fix assorted small typos: dead __str__, unformatted messages, preview min-height, tooltip - processing.py: __str___ misspelled dunder never dispatched; rename to __str__ - demofusion.py: missing f-prefix printed literal {width}x{height} in user-facing error - sparse_backend.py: missing f-prefix printed literal {size} in ValueError - generation_parameters_copypaste.py: debug message typo 'Not not registered' - progressBar.ts: live-preview gallery minHeight computed from naturalWidth; mirrors the adjacent maxHeight line (naturalHeight) - settings.ts: stray } in settings-tab tooltip | 2 个月前 | |
linting Signed-off-by: Vladimir Mandic <mandic00@live.com> | 2 个月前 | |
strict ruff without exclude Co-authored-by: Copilot <copilot@github.com> Signed-off-by: Vladimir Mandic <mandic00@live.com> | 3 个月前 | |
full codespell coverage Signed-off-by: Vladimir Mandic <mandic00@live.com> | 3 个月前 | |
linting Signed-off-by: Vladimir Mandic <mandic00@live.com> | 2 个月前 | |
video model loader Signed-off-by: Vladimir Mandic <mandic00@live.com> | 1 个月前 | |
refactor imports | 6 个月前 | |
SDNQ Atten skip TE | 1 个月前 | |
cleanup logger | 6 个月前 | |
cleanup logger | 6 个月前 | |
update readme and some debug functions Signed-off-by: vladmandic <mandic00@live.com> | 5 个月前 | |
Add SDNQ attention | 2 个月前 | |
mps install torchsde Signed-off-by: Vladimir Mandic <mandic00@live.com> | 2 个月前 | |
run enso extension installer after clone and update After deletion and re-clone, enso.install() runs after install_extensions() has already passed, so install.py was never triggered and the frontend was not built. | 5 个月前 | |
Typing cleanup | 3 个月前 | |
fix(samplers): report validation failures without backtrace Sampler capability gates raised plain ValueError when schedulers_fallback is disabled, so the API middleware and the gradio call wrapper printed a full backtrace for an expected outcome. Add errors.ValidationError, raise it from the gates, and report it message-only in errors.display; the UI error box and the API error response already carry the message. | 2 个月前 | |
handle invalid extension Signed-off-by: Vladimir Mandic <mandic00@live.com> | 2 个月前 | |
fix(lora): apply te networks before encode and honor lora_apply_te Network activation ran after prompt encoding, so text encoder lora weights never affected embeds on the first generation and the stale result was then served from the embed cache. The trailing unfiltered activate in network_load also overrode the te exclude filter, so the lora_apply_te setting was never honored. - parse and activate networks in process_base before pipeline args are built - activate_filtered gates text encoder components on per-request or global lora_apply_te; used by base, hires, detailer and faceid call sites - network_load accepts activate=False for callers that run their own deactivate/activate sequence with include/exclude - network_activate walks excluded components in restore-only mode so a filtered text encoder reverts to backup instead of keeping stale deltas - loaded_loras cache is single-entry since per-filter entries go stale when the setting toggles - prompt embed cache key includes the effective lora_apply_te value | 1 个月前 | |
fix ui/misc: None desc crash, ratio/tile/lora split guards, parse_metadtaa typo, moondream type guard Co-Authored-By: Claude <noreply@anthropic.com> | 2 个月前 | |
full codespell coverage Signed-off-by: Vladimir Mandic <mandic00@live.com> | 3 个月前 | |
Fix assorted small typos: dead __str__, unformatted messages, preview min-height, tooltip - processing.py: __str___ misspelled dunder never dispatched; rename to __str__ - demofusion.py: missing f-prefix printed literal {width}x{height} in user-facing error - sparse_backend.py: missing f-prefix printed literal {size} in ValueError - generation_parameters_copypaste.py: debug message typo 'Not not registered' - progressBar.ts: live-preview gallery minHeight computed from naturalWidth; mirrors the adjacent maxHeight line (naturalHeight) - settings.ts: stray } in settings-tab tooltip | 2 个月前 | |
gradio temp files guard Signed-off-by: Vladimir Mandic <mandic00@live.com> | 3 个月前 | |
gradio temp files guard Signed-off-by: Vladimir Mandic <mandic00@live.com> | 3 个月前 | |
validate reference, lint and update diffusers Signed-off-by: vladmandic <mandic00@live.com> | 4 个月前 | |
remove process preview Co-authored-by: Copilot <copilot@github.com> Signed-off-by: Vladimir Mandic <mandic00@live.com> | 4 个月前 | |
Revert "much stricter ruff linting" This reverts commit 310dbf1574568a0e5efb999482aa80a61e96b696. | 3 个月前 | |
minor fixes from audit Signed-off-by: Vladimir Mandic <mandic00@live.com> | 2 个月前 | |
fix infotext Signed-off-by: Vladimir Mandic <mandic00@live.com> | 1 个月前 | |
ruff lint | 6 个月前 | |
refactor detailer Signed-off-by: Vladimir Mandic <mandic00@live.com> | 2 个月前 | |
model metadata handle invalid keys Signed-off-by: Vladimir Mandic <mandic00@live.com> | 2 个月前 | |
unified logger | 6 个月前 | |
pin correct tokenizers Signed-off-by: Vladimir Mandic <mandic00@live.com> | 2 个月前 | |
cleanup logger | 6 个月前 | |
fixes Signed-off-by: Vladimir Mandic <mandic00@live.com> | 2 个月前 | |
fix(mask): guard zero-area divide in auto-mask Edge mode In get_mask() Edge mode, largest_size is set to 0 when there are no contours, but the luminance loop is also entered when contours exist whose largest has zero area (e.g. thin 1px / degenerate features whose cv2.contourArea is 0). That path divides by largest_size and raises ZeroDivisionError, crashing auto-masking. Guard the divide: when largest_size is 0 the scaled luminance is 0, so the loop breaks on the first iteration and yields an empty mask (the sensible no-region result). Co-Authored-By: Claude <noreply@anthropic.com> | 2 个月前 | |
update installer usage | 6 个月前 | |
fix model/vae: wrong quant_config var, getattr guards, remote-vae latent_copy shadowing + response handling Co-Authored-By: Claude <noreply@anthropic.com> | 2 个月前 | |
unified logger | 6 个月前 | |
refactor uv support Signed-off-by: Vladimir Mandic <mandic00@live.com> | 3 个月前 | |
feat(model): detect header metadata and nvfp4 in the model probe detect_quant now reads _quantization_metadata as the authoritative quant source when present, and resolves marker-file formats from the marked layers' stored weight dtypes instead of file-wide dtype voting, which mislabeled fp8 files carrying extra uint8 tensors and had no nvfp4 mapping at all. Schema bump so cached probe entries refresh. | 1 个月前 | |
debounce width/height updates Signed-off-by: Vladimir Mandic <mandic00@live.com> | 2 个月前 | |
fix model/vae: wrong quant_config var, getattr guards, remote-vae latent_copy shadowing + response handling Co-Authored-By: Claude <noreply@anthropic.com> | 2 个月前 | |
cleanup logger | 6 个月前 | |
add boogu Signed-off-by: Vladimir Mandic <mandic00@live.com> | 2 个月前 | |
fix(modelloader): convert ModelCardData to dict before saving model_info.json | 3 个月前 | |
add boogu Signed-off-by: Vladimir Mandic <mandic00@live.com> | 2 个月前 | |
fix non-english git and add faster stat Signed-off-by: Vladimir Mandic <mandic00@live.com> | 1 个月前 | |
cleanup logger | 6 个月前 | |
cleanup logger | 6 个月前 | |
modernize typing | 6 个月前 | |
unified logger | 6 个月前 | |
fix(api): set compatibility options without crashing clip_skip and the uni_pc_* opts live in opts.data without an OptionInfo in data_labels (compatibility_opts). Options.set() read data_labels[key].onchange unconditionally, so setting clip_skip via /sdapi/v1/options raised KeyError and returned 500; the override_settings restore path had the same unguarded data_labels[k] access for falsy-valued compat opts. Guard the onchange lookup and read the stored value via getattr(opts, k), which already falls back through data then data_labels. | 2 个月前 | |
cleanup logger | 6 个月前 | |
add wildcards to image metadata and note separate template from prompt field Signed-off-by: Vladimir Mandic <mandic00@live.com> | 3 个月前 | |
chore: rename probe cache files to signatures.json and civitai.json | 1 个月前 | |
ruff lint | 6 个月前 | |
optimize sdnq quant-on-load and add platform stats Signed-off-by: Vladimir Mandic <mandic00@live.com> | 2 个月前 | |
optimize sdnq quant-on-load and add platform stats Signed-off-by: Vladimir Mandic <mandic00@live.com> | 2 个月前 | |
feat(api): add detailer postprocess script and /sdapi/v1/detail endpoint Surface YoloRestorer.restore() as a standalone operation: a Detailer postprocessing script in the Process tab and a thin /sdapi/v1/detail endpoint, neither requiring a base generation pass. - modules/postprocess/yolo.py: YoloRestorer.make_processing() builds the synthetic Img2Img processing object both entry points feed to restore(), resolving the seed so the inpaint passes are reproducible - modules/api/process.py: post_detail handler exposes the full detailer parameter set and returns the detailed image plus optional annotations as base64 - scripts/postprocessing_detailer.py: reuses shared.yolo.ui('extras') and runs through make_processing() - modules/postprocessing.py: run_extras takes a per-script script_args dict, also letting the extras API drive other scripts such as Remove background; omitting it leaves existing callers unchanged - modules/api/models.py: ReqDetail / ResDetail - modules/processing_info.py: guard create_infotext's Image/Hires CFG reporting against an unset (None) cfg_image, matching the is-not-None checks the other cfg_image readers use; the detailer inpaint pass runs with it unset - test/test-detailer-api.py: covers both paths; effect tests measure the diff inside the detected region with extreme isolated parameter values, and the suite disables model quantization for the run and restores the original settings afterward | 3 个月前 | |
refactor detailer Signed-off-by: Vladimir Mandic <mandic00@live.com> | 2 个月前 | |
fix(lora): apply te networks before encode and honor lora_apply_te Network activation ran after prompt encoding, so text encoder lora weights never affected embeds on the first generation and the stale result was then served from the embed cache. The trailing unfiltered activate in network_load also overrode the te exclude filter, so the lora_apply_te setting was never honored. - parse and activate networks in process_base before pipeline args are built - activate_filtered gates text encoder components on per-request or global lora_apply_te; used by base, hires, detailer and faceid call sites - network_load accepts activate=False for callers that run their own deactivate/activate sequence with include/exclude - network_activate walks excluded components in restore-only mode so a filtered text encoder reverts to backup instead of keeping stale deltas - loaded_loras cache is single-entry since per-filter entries go stale when the setting toggles - prompt embed cache key includes the effective lora_apply_te value | 1 个月前 | |
schedulers fix zero-sigma final-step Signed-off-by: Vladimir Mandic <mandic00@live.com> | 2 个月前 | |
fix: dead hasattr-on-dict path, double video-vae normalization, undefined self.device, hyimage if/elif, attention-mask attr typo, wrong face logged Co-Authored-By: Claude <noreply@anthropic.com> | 2 个月前 | |
add hidream-o1 Co-authored-by: Copilot <copilot@github.com> Signed-off-by: Vladimir Mandic <mandic00@live.com> | 3 个月前 | |
fix(lora): apply te networks before encode and honor lora_apply_te Network activation ran after prompt encoding, so text encoder lora weights never affected embeds on the first generation and the stale result was then served from the embed cache. The trailing unfiltered activate in network_load also overrode the te exclude filter, so the lora_apply_te setting was never honored. - parse and activate networks in process_base before pipeline args are built - activate_filtered gates text encoder components on per-request or global lora_apply_te; used by base, hires, detailer and faceid call sites - network_load accepts activate=False for callers that run their own deactivate/activate sequence with include/exclude - network_activate walks excluded components in restore-only mode so a filtered text encoder reverts to backup instead of keeping stale deltas - loaded_loras cache is single-entry since per-filter entries go stale when the setting toggles - prompt embed cache key includes the effective lora_apply_te value | 1 个月前 | |
fix lut Co-authored-by: Copilot <copilot@github.com> Signed-off-by: Vladimir Mandic <mandic00@live.com> | 3 个月前 | |
feat(samplers): group sampler dropdown into labeled sections Reorder samplers_data_diffusers into recognizable solver-family groups (Euler, DPM/DPM++, UniPC/DEIS, Heun/KDPM2, ER-SDE, Classic, Distilled, Misc), each ending with its FlowMatch variants, and Res4Lyf as a fenced experimental section, so the dropdown is scannable. Dividers are SamplerData sentinels with U+2500 names: create_sampler keeps the current scheduler when one is selected, get_sampler_name falls back to Default, set_samplers and validate_sampler_name exclude them, and a visible_samplers() helper drops them from the xyz axes, detailer, and folder pickers. The main and refine dropdowns render them as section labels. No sampler is removed or renamed, so saved infotexts, styles, and API calls keep resolving. | 2 个月前 | |
fix batch list padding in create_infotext corrupting infotext indices padding loops used insert(0) which shifts existing prompt/seed entries to higher indices, so subsequent index lookups - including the caller's own p.prompts[i]/p.seeds[i] reads on the same aliased lists - return values belonging to a different image; append preserves existing index-to-entry mapping and pads missing slots at the end https://claude.ai/code/session_01UZT4ypkGQH5fzJ89dXPFYp | 2 个月前 | |
fix processing: init_images hasattr typo, empty-list guards, dead timer profile branch Co-Authored-By: Claude <noreply@anthropic.com> | 2 个月前 | |
schedulers fix zero-sigma final-step Signed-off-by: Vladimir Mandic <mandic00@live.com> | 2 个月前 | |
feat(video): video_interpolate as execution-time stage Promote RIFE interpolation from a save-time kwarg to a real stage of the processing pipeline so per-frame work (detailer, color correction, postprocess scripts) operates on source-rate frames and the inflated stream becomes the saved output. - new modules/processing_video.py with apply_video_interpolation, interpolation_factor, expand_infotexts; PIL/tensor/numpy dispatch - video_interpolate, video_interpolate_scale, video_interpolated fields on StableDiffusionProcessingVideo - process_images_inner runs the helper after the batch loop and inflates infotexts in lockstep - save_video in modules/video.py and modules/video_models/video_save.py short-circuit re-interpolation when p.video_interpolated is set; the user-facing kwarg still flows into metadata | 4 个月前 | |
fix preview mapping Signed-off-by: Vladimir Mandic <mandic00@live.com> | 2 个月前 | |
fix processing: init_images hasattr typo, empty-list guards, dead timer profile branch Co-Authored-By: Claude <noreply@anthropic.com> | 2 个月前 | |
fix(lora): apply te networks before encode and honor lora_apply_te Network activation ran after prompt encoding, so text encoder lora weights never affected embeds on the first generation and the stale result was then served from the embed cache. The trailing unfiltered activate in network_load also overrode the te exclude filter, so the lora_apply_te setting was never honored. - parse and activate networks in process_base before pipeline args are built - activate_filtered gates text encoder components on per-request or global lora_apply_te; used by base, hires, detailer and faceid call sites - network_load accepts activate=False for callers that run their own deactivate/activate sequence with include/exclude - network_activate walks excluded components in restore-only mode so a filtered text encoder reverts to backup instead of keeping stale deltas - loaded_loras cache is single-entry since per-filter entries go stale when the setting toggles - prompt embed cache key includes the effective lora_apply_te value | 1 个月前 | |
fix(prompt): restore textual inversion on clip-skip>=2 transformers 5.6 flattened CLIPTextModel, removing the .text_model wrapper that compel_hijack and the xhinker parser dereference on the normalized clip-skip path. On SD1.5 at clip-skip >= 2 this raised AttributeError, which processing_prompt caught and silently fell back to fixed-attention encoding, dropping textual inversion and prompt weighting. Resolve the submodule via getattr(te, 'text_model', te), correct for flattened CLIPTextModel, CLIPTextModelWithProjection (still nested), and transformers < 5.6. | 2 个月前 | |
rocm: detect hipBLASLt lazy library in per-arch subfolder and zlib form | 1 个月前 | |
refactor imports | 6 个月前 | |
linting Signed-off-by: Vladimir Mandic <mandic00@live.com> | 10 个月前 | |
Move functions to match new execution order | 4 个月前 | |
cleanup logging Signed-off-by: Vladimir Mandic <mandic00@live.com> | 3 个月前 | |
strict ruff without exclude Co-authored-by: Copilot <copilot@github.com> Signed-off-by: Vladimir Mandic <mandic00@live.com> | 3 个月前 | |
modernize typing | 6 个月前 | |
Fix per-script timing always recording zero ScriptSummary.record reset the reference timestamp immediately before measuring against it, so every recorded duration was ~0 and report() never logged script timings. Measure against the previous timestamp, then advance it. https: //claude.ai/code/session_014QWKWgKvMevcuvfCnsYoT2 Co-Authored-By: Claude <noreply@anthropic.com> | 2 个月前 | |
fix processing: init_images hasattr typo, empty-list guards, dead timer profile branch Co-Authored-By: Claude <noreply@anthropic.com> | 2 个月前 | |
model metadata handle invalid keys Signed-off-by: Vladimir Mandic <mandic00@live.com> | 2 个月前 | |
add boogu Signed-off-by: Vladimir Mandic <mandic00@live.com> | 2 个月前 | |
initial pyright lint Signed-off-by: Vladimir Mandic <mandic00@live.com> | 6 个月前 | |
fix hijack accelerate Signed-off-by: Vladimir Mandic <mandic00@live.com> | 2 个月前 | |
modernize typing | 6 个月前 | |
cleanup Signed-off-by: vladmandic <mandic00@live.com> | 5 个月前 | |
update hf hijack, simplify anima loader, add few more models Signed-off-by: Vladimir Mandic <mandic00@live.com> | 2 个月前 | |
fix(hypertile): correct width/height mapping in hires set_resolution hr_upscale_to_x is the width dimension and hr_upscale_to_y the height (see processing_class.py); set_resolution assigned them transposed, producing swapped HyperTile geometry on non-square hires passes. Co-Authored-By: Claude <noreply@anthropic.com> | 2 个月前 | |
unified logger | 6 个月前 | |
add ideogram prequant and set default schedule Signed-off-by: Vladimir Mandic <mandic00@live.com> | 2 个月前 | |
add boogu Signed-off-by: Vladimir Mandic <mandic00@live.com> | 2 个月前 | |
complete refactor javascript to typescript and reorg frontend files and folders Signed-off-by: Vladimir Mandic <mandic00@live.com> | 3 个月前 | |
modernize typing | 6 个月前 | |
add boogu Signed-off-by: Vladimir Mandic <mandic00@live.com> | 2 个月前 | |
add joy-image-edit-plus Signed-off-by: Vladimir Mandic <mandic00@live.com> | 1 个月前 | |
add experimental pruna Signed-off-by: Vladimir Mandic <mandic00@live.com> | 2 个月前 | |
video model loader Signed-off-by: Vladimir Mandic <mandic00@live.com> | 1 个月前 | |
full ui-settings refactor Signed-off-by: Vladimir Mandic <mandic00@live.com> | 1 年前 | |
SDNQ skip quant if numel is less than 16384 and add nn.Embedding to no spilt modules on offload | 1 个月前 | |
RUF013 updates and typing update | 5 个月前 | |
feat(samplers): group sampler dropdown into labeled sections Reorder samplers_data_diffusers into recognizable solver-family groups (Euler, DPM/DPM++, UniPC/DEIS, Heun/KDPM2, ER-SDE, Classic, Distilled, Misc), each ending with its FlowMatch variants, and Res4Lyf as a fenced experimental section, so the dropdown is scannable. Dividers are SamplerData sentinels with U+2500 names: create_sampler keeps the current scheduler when one is selected, get_sampler_name falls back to Default, set_samplers and validate_sampler_name exclude them, and a visible_samplers() helper drops them from the xyz axes, detailer, and folder pickers. The main and refine dropdowns render them as section labels. No sampler is removed or renamed, so saved infotexts, styles, and API calls keep resolving. | 2 个月前 | |
feat(krea2): add Krea 2 (K2) image model support Krea 2 is a 12.9B single-stream flow-matching DiT trained from scratch, using a Qwen3-VL-4B text encoder and the Qwen-Image VAE. The transformer is vendored as a diffusers ModelMixin whose module tree mirrors the checkpoint, so weights load with no key conversion; the pipeline ports the reference encode, flow-matching denoise, and VAE decode. The text encoder is shared at runtime via the existing dedup registry, so Base and Turbo reuse one Qwen3-VL-4B copy. Covers text-to-image, image-to-image, native LoRA, and the single-file UNET override. Also completes SD.Next's partial Qwen-Image VAE support (5D decode input and TAESD preview mapping) that K2 shares. | 2 个月前 | |
feat(samplers): add DPM++ 3M SDE preset DPMSolverMultistepScheduler implements 3rd-order sde-dpmsolver++, but the preset grid stopped at order 2 on the SDE row. Add the order-3 variant next to DPM++ SDE and DPM++ 2M SDE to complete it. | 2 个月前 | |
RUF013 updates and typing update | 5 个月前 | |
feat(model): secondary unet override slot and ideogram4 native loading One UNET override cannot serve dual-transformer arches: ideogram4 conditional/unconditional and wan combined-stage experts need separate files, and previously a single override landed on both experts. - sd_unet_secondary option with per-slot tracking, consumed-state sync, arch-change reset, and incompatible-override fallback - dropdown renders beside the primary, follows it into quicksettings, and is visible only for dual-transformer model types - ideogram4 native single-file spec with a quant-aware fused-qkv converter; such converters run before comfy_quant detection via TransformerSpec.converter_handles_quant - quicksettings render in configured order (sort keyed on the option object and always fell back to alphabetical) - post-load dtype warning skips quantized transformers | 1 个月前 | |
vae_scale_factor override Signed-off-by: Vladimir Mandic <mandic00@live.com> | 2 个月前 | |
lint fix Signed-off-by: Vladimir Mandic <mandic00@live.com> | 3 个月前 | |
refactor detailer Signed-off-by: Vladimir Mandic <mandic00@live.com> | 2 个月前 | |
Add SDNQ attention | 2 个月前 | |
unified logger | 6 个月前 | |
refactor(ui): drop quicksettings companion machinery for secondary unet The secondary unet loads from the networks panel toggle or the settings page; companion ordering and dynamic visibility pushes in ui_settings.py are not worth their footprint for one setting. | 1 个月前 | |
- all guidance values set to -1 to enable using model defaults - log default values used if not overriden by user - rename inconsistent guidance variables: p.cfg_scale p.image_cfg_scale -> p.cfg_image p.diffusers_guidance_rescale -> p.cfg_rescale p.pag_scale -> p.cfg_true p.pag_adaptive -> p.cfg_adaptive Signed-off-by: Vladimir Mandic <mandic00@live.com> | 3 个月前 | |
add wildcards to image metadata and note separate template from prompt field Signed-off-by: Vladimir Mandic <mandic00@live.com> | 3 个月前 | |
add ideogram4 Signed-off-by: Vladimir Mandic <mandic00@live.com> | 3 个月前 | |
modernize typing | 6 个月前 | |
fix(embeddings): restore multi-vector textual inversion expansion transformers 5 defaults tokenizer.add_tokens() to normalized=True, so the CLIP tokenizer lowercases added embedding names. tokenizer.tokenize() then returns the lowercased surface, maybe_convert_prompt never matches mixed-case names in added_tokens_encoder, and multi-vector expansion is skipped. Every mixed-case multi-vector embedding collapsed to its first vector and looked ignored. Add embedding tokens as AddedToken(name, normalized=False) so they stay case-sensitive and tokenize() surfaces them verbatim. Valid on transformers 4.x and 5.x; convert_tokens_to_ids and encoding are unaffected. | 2 个月前 | |
add legacy notice Signed-off-by: Vladimir Mandic <mandic00@live.com> | 2 个月前 | |
cleanup performance tags Signed-off-by: Vladimir Mandic <mandic00@live.com> | 2 个月前 | |
cleanup logging and update requirements Signed-off-by: vladmandic <mandic00@live.com> | 5 个月前 | |
cleanup logger | 6 个月前 | |
- all guidance values set to -1 to enable using model defaults - log default values used if not overriden by user - rename inconsistent guidance variables: p.cfg_scale p.image_cfg_scale -> p.cfg_image p.diffusers_guidance_rescale -> p.cfg_rescale p.pag_scale -> p.cfg_true p.pag_adaptive -> p.cfg_adaptive Signed-off-by: Vladimir Mandic <mandic00@live.com> | 3 个月前 | |
networks load/download model track progress Co-authored-by: Copilot <copilot@github.com> Signed-off-by: Vladimir Mandic <mandic00@live.com> | 3 个月前 | |
css updates Signed-off-by: Vladimir Mandic <mandic00@live.com> | 1 个月前 | |
fix gallery delete and save Signed-off-by: Vladimir Mandic <mandic00@live.com> | 1 个月前 | |
full codespell coverage Signed-off-by: Vladimir Mandic <mandic00@live.com> | 3 个月前 | |
refactor detailer Signed-off-by: Vladimir Mandic <mandic00@live.com> | 2 个月前 | |
fix(control): remove inert DWPose (Legacy) pose-model selector The DWPose (Legacy) "Pose Model" (Tiny/Medium/Large) radio wrote to config['DWPose (Legacy)']['model'], but the rtmlib backend (RtmlibPoseDetector.from_pretrained) only takes the mode string and always loads RTMPose 'lightweight' - it never reads that key (the only config[...]['model'] readers are the SegmentAnything branch). The selector was a no-op left over from the removed mmpose DWPose processor (see #4956). Remove the radio, its dead 'model' config key (both the module-level and delay_load_config dicts), and its update_settings slot; reindex the remaining settings 18..27. The settings list stays 1:1 with the UI controls (re-derived the append-order -> update_settings map to confirm). No runtime behavior change. The alternative - keep the capability by mapping Tiny/Medium/Large onto rtmlib's mode - is intentionally not taken here since it would change which model loads. Co-Authored-By: Claude <noreply@anthropic.com> | 2 个月前 | |
update masking Signed-off-by: Vladimir Mandic <mandic00@live.com> | 3 个月前 | |
refactor(ui): drop quicksettings companion machinery for secondary unet The secondary unet loads from the networks panel toggle or the settings page; companion ordering and dynamic visibility pushes in ui_settings.py are not worth their footprint for one setting. | 1 个月前 | |
pylint | 6 个月前 | |
handle invalid extension Signed-off-by: Vladimir Mandic <mandic00@live.com> | 2 个月前 | |
feat(ui): load unet into secondary slot from networks panel The refine toggle doubles as the slot selector on the UNet/DiT page: with it active a card click sets sd_unet_secondary instead of sd_unet, mirroring base/refiner selection on the model page. | 1 个月前 | |
fix(networks): keep reference mtime a datetime for date sort strptime was handed the stat datetime instead of the date string, so it always failed and left a raw string mtime; the Date sort then crashed on mixed str/datetime items. Parse the date string, falling back to stat mtime. | 2 个月前 | |
pylint | 6 个月前 | |
feat(krea2): add Krea 2 (K2) image model support Krea 2 is a 12.9B single-stream flow-matching DiT trained from scratch, using a Qwen3-VL-4B text encoder and the Qwen-Image VAE. The transformer is vendored as a diffusers ModelMixin whose module tree mirrors the checkpoint, so weights load with no key conversion; the pipeline ports the reference encode, flow-matching denoise, and VAE decode. The text encoder is shared at runtime via the existing dedup registry, so Base and Turbo reuse one Qwen3-VL-4B copy. Covers text-to-image, image-to-image, native LoRA, and the single-file UNET override. Also completes SD.Next's partial Qwen-Image VAE support (5D decode input and TAESD preview mapping) that K2 shares. | 2 个月前 | |
fix ui/misc: None desc crash, ratio/tile/lora split guards, parse_metadtaa typo, moondream type guard Co-Authored-By: Claude <noreply@anthropic.com> | 2 个月前 | |
cleanup references to p.sd_model Signed-off-by: Vladimir Mandic <mandic00@live.com> | 6 个月前 | |
fix non-english git and add faster stat Signed-off-by: Vladimir Mandic <mandic00@live.com> | 1 个月前 | |
cleanup logger | 6 个月前 | |
feat(autocomplete): aliases, extra-networks, translations, baseline fixes - tag aliases (optional 4-tuple) and translations (companion file) - lora and wildcard triggers; embeddings fold into tag-mode results - new GET /sdapi/v1/wildcards endpoint - LRU cap, IME safety, multi-textarea safety, modifier-key pass-through - manifest-retry and config-bridge hardening - tags-fetch: alias/translation harvesters with authenticated rate limits | 4 个月前 | |
ui fixes Signed-off-by: Vladimir Mandic <mandic00@live.com> | 3 个月前 | |
- all guidance values set to -1 to enable using model defaults - log default values used if not overriden by user - rename inconsistent guidance variables: p.cfg_scale p.image_cfg_scale -> p.cfg_image p.diffusers_guidance_rescale -> p.cfg_rescale p.pag_scale -> p.cfg_true p.pag_adaptive -> p.cfg_adaptive Signed-off-by: Vladimir Mandic <mandic00@live.com> | 3 个月前 | |
show_progress requires "full", "minimal", or "hidden" | 9 个月前 | |
refactor detailer Signed-off-by: Vladimir Mandic <mandic00@live.com> | 2 个月前 | |
additional exports Signed-off-by: Vladimir Mandic <mandic00@live.com> | 3 个月前 | |
refactor imports | 6 个月前 | |
improve ui models Signed-off-by: Vladimir Mandic <mandic00@live.com> | 1 个月前 | |
cleanup Signed-off-by: Vladimir Mandic <mandic00@live.com> | 3 个月前 | |
cleanup logger | 6 个月前 | |
fix save style without name | 2 年前 | |
Merge pull request #4971 from vladmandic/fix/resolution-ar-controls fix(ui): debounce aspect-ratio linked width/height controls | 2 个月前 | |
feat(ui): load unet into secondary slot from networks panel The refine toggle doubles as the slot selector on the UNet/DiT page: with it active a card click sets sd_unet_secondary instead of sd_unet, mirroring base/refiner selection on the model page. | 1 个月前 | |
gallery add thumbnail size slider Co-authored-by: Copilot <copilot@github.com> Signed-off-by: Vladimir Mandic <mandic00@live.com> | 4 个月前 | |
refactor detailer Signed-off-by: Vladimir Mandic <mandic00@live.com> | 2 个月前 | |
feat(video): add thumbnail save toggle | 3 个月前 | |
RUF013 updates and formatting | 5 个月前 | |
Fix url formatting in update module | 9 个月前 | |
RUF013 updates and import updates | 5 个月前 | |
PIL Image.Image type fix | 4 个月前 | |
ty lint Signed-off-by: Vladimir Mandic <mandic00@live.com> | 3 个月前 | |
PIL Image.Image type fix | 4 个月前 | |
PIL Image.Image type fix | 4 个月前 | |
fix ui/misc: None desc crash, ratio/tile/lora split guards, parse_metadtaa typo, moondream type guard Co-Authored-By: Claude <noreply@anthropic.com> | 2 个月前 | |
unified logger | 6 个月前 | |
unified logger | 6 个月前 |