Public/automatic
1
0
Fork 0
mirror of https://github.com/vladmandic/automatic synced 2026-09-10 23:08:43 +02:00
Code Issues Packages Projects Releases Wiki Activity
Files
443a73b740fd591db516f76d16735802cb110075
automatic/modules
T
History
CalamitousFelicitousness 443a73b740 refactor(caption): code review fixes for offload, inference, and maintainability 
Comprehensive review of modules/caption/ addressing memory management,
consistency, and code quality:

Inference correctness:
- Add devices.inference_context() to _qwen(), _smol(), _sa2() handlers
- Remove redundant @torch.no_grad() decorator from joycaption predict()
- Remove dead dtype=torch.bfloat16 kwarg from Florence loader

Memory management:
- Bound moondream3 image cache with LRU eviction (max 8 entries)
- Replace fragile id(image) cache keys with content-based md5 hash
- Add devices.torch_gc() after model loading in deepseek
- Move deepbooru model to CPU before dropping reference on unload
- Add external handler delegation to VQA.unload() (moondream3,
  joycaption, joytag, deepseek)
- Protect batch offload mutation with try/finally

Code deduplication:
- Extract strip_think_xml_tags() shared helper for Qwen/Gemma/SmolVLM
- Extract save_tags_to_file() into tagger.py from deepbooru and
  waifudiffusion

Documentation and clarity:
- Document deepseek global monkey-patches (LlamaFlashAttention2, attrdict)
- Document Florence task="task" as intentional design choice
- Add vendored-code comment to joytag.py
- Document openclip direct .to() usage vs sd_models.move_model
- Comment model.eval() calls that are required (trust_remote_code,
  custom loaders) vs removed where redundant (standard from_pretrained)

API robustness:
- Add HTTP 422 error response for VQA caption error strings in API
  endpoints (post_vqa, _dispatch_vlm)
2026-02-11 02:48:11 +00:00
..
apg
fix gallery save/delete
2026-01-20 09:53:20 +01:00
api
refactor(caption): code review fixes for offload, inference, and maintainability
2026-02-11 02:48:11 +00:00
ben2
integrate kanvas with server-side masking
2025-11-15 11:56:56 -05:00
caption
refactor(caption): code review fixes for offload, inference, and maintainability
2026-02-11 02:48:11 +00:00
cfgzero
lint formatting
2025-07-05 17:13:31 -04:00
civitai
update base models
2025-12-30 18:36:02 +01:00
control
cleanup todo
2026-01-19 11:10:05 +01:00
dml
add builtin framepack
2025-07-08 15:47:07 -04:00
face
Type standardization in processing_class
2026-01-21 16:35:19 -08:00
facelib
cleanup todo
2026-01-19 11:10:05 +01:00
flash_attn_triton_amd
cleanup todo
2026-01-19 11:10:05 +01:00
framepack
move vae to subfolder
2026-01-30 11:34:25 +01:00
ggml
Fix importlib.metadata
2025-11-27 14:47:56 -08:00
hidiffusion
fix gallery save/delete
2026-01-20 09:53:20 +01:00
intel
cleanup
2026-01-10 20:48:36 +03:00
linfusion
add kanvas integration
2025-11-09 15:41:20 -05:00
lora
add --remote cmd opt
2026-02-07 10:17:32 +00:00
ltx
fix/add element names for some video controls that were duplicate/missing
2026-01-21 19:53:11 -05:00
merging
autodetect v-pred
2025-07-28 19:41:39 -04:00
onnx_impl
fix gallery save/delete
2026-01-20 09:53:20 +01:00
pag
Fix if cls is None
2026-01-21 18:16:47 -08:00
postprocess
fix anima
2026-02-05 16:55:06 +01:00
ras
cleanup todo
2026-01-19 11:10:05 +01:00
res4lyf
add sampler api endpoints
2026-02-04 13:08:31 +01:00
rife
add lbm background replace with relightining
2025-07-04 15:33:16 -04:00
schedulers
add sampler api endpoints
2026-02-04 13:08:31 +01:00
sdnq
SDNQ update CosmosTransformer3DModel keys
2026-02-04 10:19:23 +03:00
seedvr
linting
2025-10-16 19:39:05 -04:00
taesd
cleanup todo
2026-01-19 11:10:05 +01:00
teacache
lint formatting
2025-07-05 17:13:31 -04:00
todo
cleanup
2025-01-10 08:25:00 -05:00
vae
fix lora unload and improve preview error handler
2026-02-07 09:07:34 +00:00
video_models
move vae to subfolder
2026-01-30 11:34:25 +01:00
attention.py
get_hip_arch_name -> get_hip_agent, use amdhip64_7.dll served within rocm package
2026-01-03 21:00:36 +09:00
cachedit.py
allow lora specifier to select model component. see changelog/wiki for docs on how to.
2025-09-23 15:25:13 -04:00
call_queue.py
improve wrap_gradio_call
2026-01-20 18:56:49 +01:00
cmd_args.py
ability to disable parts of the app
2025-11-03 19:36:55 -05:00
detailer.py
detailer support for face restorer models
2025-01-18 14:29:40 -05:00
devices_mac.py
ruff updates
2024-12-20 18:58:29 -05:00
devices.py
cleanup logging
2026-01-16 10:36:02 +01:00
errorlimiter.py
Move to modules
2026-01-24 02:16:05 -08:00
errors.py
log captured exceptions
2026-01-24 13:57:13 +01:00
extensions.py
Merge pull request #4507 from awsr/ext-table
2026-01-01 10:28:52 +01:00
extra_networks.py
allow different lora in hires
2026-02-09 22:31:00 +01:00
extras.py
upadte qwen pruning and allow hf models in subfolders
2025-10-04 15:49:20 -04:00
face_restoration.py
…
files_cache.py
add SD_NO_CACHE env variable
2024-12-11 12:32:34 -05:00
generation_parameters_copypaste.py
fix metadata save and temp file handler
2026-01-28 19:28:49 +01:00
gr_hijack.py
kanvas bindings
2025-11-07 12:21:48 -05:00
gr_tempdir.py
fix metadata save and temp file handler
2026-01-28 19:28:49 +01:00
hashes.py
relocate all jsons to data
2026-01-24 13:54:40 +01:00
history.py
cleanup todo
2026-01-19 11:10:05 +01:00
images_grid.py
guard against images list and avoid pipeline switches
2025-10-28 14:38:35 -04:00
images_namegen.py
fix metadata save and temp file handler
2026-01-28 19:28:49 +01:00
images_resize.py
reduce requirements
2025-11-12 19:24:47 -05:00
images.py
Revert "Mark public re-exports"
2026-02-09 18:30:18 +01:00
img2img.py
feat(settings): add base path support for output folders
2026-01-16 16:24:05 +00:00
infotext_utils.py
fix ansi controle output and add model locking
2025-07-06 12:50:09 -04:00
infotext.py
update override settings
2025-10-30 14:01:32 -04:00
ipadapter.py
Update typing
2026-01-21 18:15:20 -08:00
json_helpers.py
relocate all jsons to data
2026-01-24 13:54:40 +01:00
lama.py
styles support parsed and upparsed save and apply
2025-02-11 11:33:37 -05:00
loader.py
res4lyf epsilon validated
2026-01-31 12:37:17 +00:00
localization.py
…
masking.py
cleanup todo
2026-01-19 11:10:05 +01:00
memmon.py
update memmon
2025-04-21 20:20:04 -04:00
memstats.py
lint
2025-11-17 20:12:47 -05:00
migrate.py
relocate all jsons to data
2026-01-24 13:54:40 +01:00
mit_nunchaku.py
refactor: update nunchaku repo URLs and version handling
2026-02-07 22:27:05 +00:00
model_quant.py
feat: add Nunchaku group to reference
2026-02-07 22:27:05 +00:00
model_te.py
fix loading custom t5
2025-08-13 16:06:25 -04:00
model_tools.py
add new optimum-quanto on-the-fly and simplify quantization loading
2025-03-16 21:45:05 -04:00
modeldata.py
fix: Anima pipeline detection, custom module loading, and model type
2026-02-02 00:44:51 +00:00
modelloader.py
allow unsafe ssl context for downloads
2026-01-19 17:36:01 +01:00
models_hf.py
fix hf downloader
2025-10-17 13:42:28 -04:00
modelstats.py
upadte qwen pruning and allow hf models in subfolders
2025-10-04 15:49:20 -04:00
modular_guiders.py
add hunyuanimage-2.1
2025-10-24 10:29:58 -04:00
modular.py
add hunyuanimage-2.1
2025-10-24 10:29:58 -04:00
olive_script.py
…
options_handler.py
Rewrite to satisfy type checker
2026-01-13 16:44:54 -08:00
options.py
move vae to subfolder
2026-01-30 11:34:25 +01:00
para_attention.py
major refactor: remove backend original
2025-07-05 13:16:46 -04:00
patches.py
…
paths_internal.py
change default hfcache folder
2025-08-13 13:04:14 -04:00
paths.py
fix metadata save and temp file handler
2026-01-28 19:28:49 +01:00
postprocessing.py
feat(settings): add base path support for output folders
2026-01-16 16:24:05 +00:00
processing_args.py
feat: add Anima (Cosmos-Predict-2B variant) pipeline support
2026-02-02 00:44:51 +00:00
processing_callbacks.py
lint
2026-01-16 08:56:30 +01:00
processing_class.py
Type standardization in processing_class
2026-01-21 16:35:19 -08:00
processing_correction.py
move vae to subfolder
2026-01-30 11:34:25 +01:00
processing_diffusers.py
allow different lora in hires
2026-02-09 22:31:00 +01:00
processing_helpers.py
use base steps as-is for non sd/sdxl models
2026-01-23 10:12:11 +01:00
processing_info.py
add scheduler class to image metadata
2026-02-04 13:20:26 +01:00
processing_prompt.py
allow unsafe ssl context for downloads
2026-01-19 17:36:01 +01:00
processing_vae.py
move vae to subfolder
2026-01-30 11:34:25 +01:00
processing.py
allow different lora in hires
2026-02-09 22:31:00 +01:00
progress.py
disabling live preview should not disable progress updates
2025-10-21 11:07:47 -04:00
prompt_parser_diffusers.py
refactor prompt set
2026-01-19 10:51:28 +01:00
prompt_parser_xhinker.py
lint and crlf
2026-01-24 10:28:46 +01:00
prompt_parser.py
lint
2025-08-02 15:28:45 -04:00
rocm_triton_windows.py
Fix rocm triton windows patch
2026-01-20 20:58:04 +08:00
rocm.py
update therock sources
2026-02-02 12:28:43 +09:00
safe.py
linting
2025-10-16 19:39:05 -04:00
script_callbacks.py
startup tracing
2025-01-08 09:35:49 -05:00
script_loading.py
fix ansi controle output and add model locking
2025-07-06 12:50:09 -04:00
scripts_auto_postprocessing.py
major refactoring of modules
2025-07-03 09:18:38 -04:00
scripts_manager.py
metadata restore to always-on scrips
2026-02-05 10:21:49 +01:00
scripts_postprocessing.py
fix postprocess
2025-11-04 08:23:45 -05:00
scripts.py
lint
2025-07-31 17:33:44 -04:00
sd_checkpoint.py
relocate all jsons to data
2026-01-24 13:54:40 +01:00
sd_detect.py
fix anima
2026-02-05 16:55:06 +01:00
sd_hijack_accelerate.py
lint
2025-08-01 18:26:57 -04:00
sd_hijack_dynamic_atten.py
add typing to sdpa hijacks
2025-09-10 04:43:57 +03:00
sd_hijack_freeu.py
css and logging tweaks
2025-07-27 08:17:41 -04:00
sd_hijack_hypertile.py
css and logging tweaks
2025-07-27 08:17:41 -04:00
sd_hijack_safetensors.py
Update runai-model-streamer logging integration
2025-12-02 02:01:51 +00:00
sd_hijack_te.py
cleanup/refactor state history
2025-09-12 16:12:45 -04:00
sd_hijack_utils.py
…
sd_hijack_vae.py
ltx2 improvements
2026-01-16 17:31:56 +00:00
sd_hijack.py
improve extensions compatibility
2025-07-05 17:42:13 -04:00
sd_models_compile.py
Unify compile Upscaler compile command
2025-09-03 18:51:21 +03:00
sd_models_utils.py
seedvt2
2025-10-12 15:35:08 -04:00
sd_models.py
one-more-offload
2026-02-04 15:05:22 +01:00
sd_modules.py
full ui-settings refactor
2025-03-30 15:04:17 -04:00
sd_offload.py
fix lora load
2026-01-21 08:45:45 +01:00
sd_samplers_common.py
fix lora unload and improve preview error handler
2026-02-07 09:07:34 +00:00
sd_samplers_diffusers.py
update schedulers and lint everything
2026-02-04 11:58:02 +01:00
sd_samplers.py
add sampler api endpoints
2026-02-04 13:08:31 +01:00
sd_te_remote.py
experiment: remote-t5
2025-08-15 10:56:36 -04:00
sd_unet.py
fix: skip Nunchaku offloading for SDXL models
2026-02-07 22:27:05 +00:00
sd_vae.py
force align width/height to vae scale factor
2026-01-05 09:42:37 +01:00
server.py
…
shared_defaults.py
get_hip_arch_name -> get_hip_agent, use amdhip64_7.dll served within rocm package
2026-01-03 21:00:36 +09:00
shared_helpers.py
refactor monolithic shared.py and separate legacy options with logging
2025-07-12 13:35:41 -04:00
shared_items.py
feat: add Anima (Cosmos-Predict-2B variant) pipeline support
2026-02-02 00:44:51 +00:00
shared_legacy.py
refactor: update imports for caption module rename
2026-02-11 02:47:41 +00:00
shared_state.py
disabling live preview should not disable progress updates
2025-10-21 11:07:47 -04:00
shared.py
fix(caption): remove dead min_length param, split Florence/PromptGen prompts, fix gaze detection
2026-02-11 02:48:11 +00:00
styles.py
weighted wildcards: corrections after automated tests
2026-02-04 19:11:05 +01:00
sub_quadratic_attention.py
…
textual_inversion.py
Move to modules
2026-01-24 02:16:05 -08:00
theme.py
relocate all jsons to data
2026-01-24 13:54:40 +01:00
timer.py
cleanup
2025-11-02 06:29:19 -05:00
token_merge.py
…
transformer_cache.py
update changelog and cleanup
2025-03-27 11:53:03 -04:00
txt2img.py
feat(settings): add base path support for output folders
2026-01-16 16:24:05 +00:00
ui_caption.py
fix(caption): remove dead min_length param, split Florence/PromptGen prompts, fix gaze detection
2026-02-11 02:48:11 +00:00
ui_common.py
refactor: update imports for caption module rename
2026-02-11 02:47:41 +00:00
ui_components.py
modernui configurable grid images size
2025-08-31 12:39:59 -04:00
ui_control_elements.py
controlnet allow processor to keep aspect-ratio for override images based on i2i or t2i resolution
2025-11-04 09:37:18 -05:00
ui_control_helpers.py
refactor: update imports for caption module rename
2026-02-11 02:47:41 +00:00
ui_control.py
refactor: update imports for caption module rename
2026-02-11 02:47:41 +00:00
ui_docs.py
show_progress requires "full", "minimal", or "hidden"
2025-11-28 02:27:19 -08:00
ui_extensions.py
relocate all jsons to data
2026-01-24 13:54:40 +01:00
ui_extra_networks_checkpoints.py
cleanup and update changelog
2026-02-09 18:20:10 +01:00
ui_extra_networks_history.py
…
ui_extra_networks_lora.py
switch lora logic and improve ui networks filter
2026-01-14 10:31:21 +01:00
ui_extra_networks_styles.py
fix detailer with lora
2025-11-09 09:14:26 -05:00
ui_extra_networks_textual_inversion.py
add wildcards to networks ui
2025-08-13 20:23:35 -04:00
ui_extra_networks_vae.py
handle missing version
2025-08-25 19:39:40 -04:00
ui_extra_networks_wildcards.py
Fix wildcard parents() termination guard
2025-09-02 01:27:58 -04:00
ui_extra_networks.py
feat: add Nunchaku group to reference
2026-02-07 22:27:05 +00:00
ui_gallery.py
Add attachment point for non-ModernUI
2026-01-04 00:37:43 -08:00
ui_guidance.py
Change the default SDXL sampler to Euler A + Linear and set Guidence scale value to 4
2025-10-24 19:58:59 +03:00
ui_history.py
show_progress requires "full", "minimal", or "hidden"
2025-11-28 02:27:19 -08:00
ui_img2img.py
refactor: update imports for caption module rename
2026-02-11 02:47:41 +00:00
ui_javascript.py
Fix incorrect variable check
2025-12-26 09:25:25 -08:00
ui_loadsave.py
lint and safeguard glm
2026-01-16 09:40:48 +01:00
ui_models_load.py
1000 papercuts
2025-05-13 21:51:33 -04:00
ui_models.py
Trim leading and trailing space from doc
2026-01-13 15:41:30 -08:00
ui_postprocessing.py
process tab send-to
2025-11-12 09:19:28 -05:00
ui_prompt_styles.py
…
ui_sections.py
refactor: update imports for caption module rename
2026-02-11 02:47:41 +00:00
ui_settings.py
Simplify options init/save/load
2026-01-10 13:27:38 -08:00
ui_symbols.py
refactor: update imports for caption module rename
2026-02-11 02:47:41 +00:00
ui_txt2img.py
show_progress requires "full", "minimal", or "hidden"
2025-11-28 02:27:19 -08:00
ui_video_vlm.py
refactor: update imports for caption module rename
2026-02-11 02:47:41 +00:00
ui_video.py
video tab add params.txt
2025-10-22 08:39:45 -04:00
ui.py
update mime-types
2025-11-14 15:47:13 -05:00
update.py
Fix url formatting in update module
2025-11-08 14:54:34 -08:00
upscaler_algo.py
add hqx and icb interpolations
2026-02-05 11:21:17 +01:00
upscaler_simple.py
add hqx and icb interpolations
2026-02-05 11:21:17 +01:00
upscaler_spandrel.py
add RealPLKSR NomosWebPhoto and AnimeSharpV2 using spandrel engine
2026-02-05 14:06:17 +01:00
upscaler_vae.py
add RealPLKSR NomosWebPhoto and AnimeSharpV2 using spandrel engine
2026-02-05 14:06:17 +01:00
upscaler.py
relocate all jsons to data
2026-01-24 13:54:40 +01:00
video.py
unified video save
2025-10-05 20:17:32 -04:00
zluda_installer.py
allow unsafe ssl context for downloads
2026-01-19 17:36:01 +01:00
zluda.py
clean up
2026-01-07 15:01:39 +09:00