da34261cc2
update
2025-10-08 20:24:54 +05:30
14cfbab078
update
2025-10-08 20:24:17 +05:30
a69b1e06fc
Merge branch 'main' into attn-refactor-blocks
2025-10-08 14:40:26 +05:30
345864eb85
fix more torch.distributed imports ( #12425 )
...
* up
* unguard.
2025-10-08 10:45:39 +05:30
35e538d46a
fix dockerfile definitions. ( #12424 )
...
* fix dockerfile definitions.
* python 3.10 slim.
* up
* up
* up
* up
* up
* revert pr_tests.yml changes
* up
* up
* reduce python version for torch 2.1.0
2025-10-08 09:46:18 +05:30
2dc31677e1
Align Flux modular more and more with Qwen modular ( #12445 )
...
* start
* fix
* up
2025-10-08 09:22:34 +05:30
1066de8c69
[Qwen LoRA training] fix bug when offloading ( #12440 )
...
* fix bug when offload and cache_latents both enabled
* fix bug when offload and cache_latents both enabled
* fix bug when offload and cache_latents both enabled
* fix bug when offload and cache_latents both enabled
* fix bug when offload and cache_latents both enabled
* fix bug when offload and cache_latents both enabled
* fix bug when offload and cache_latents both enabled
* fix bug when offload and cache_latents both enabled
* fix bug when offload and cache_latents both enabled
2025-10-07 18:27:15 +03:00
2d69bacb00
handle offload_state_dict when initing transformers models ( #12438 )
2025-10-07 13:51:20 +05:30
0974b4c606
[i18n-KO] Fix typo and update translation in ethical_guidelines.md ( #12435 )
2025-10-06 14:24:05 -07:00
cf4b97b233
[perf] Cache version checks ( #12399 )
2025-10-06 17:45:34 +02:00
7f3e9b8695
make flux ready for mellon ( #12419 )
...
* make flux ready for mellon
* up
* Apply suggestions from code review
Co-authored-by: Álvaro Somoza <asomoza@users.noreply.github.com >
---------
Co-authored-by: Álvaro Somoza <asomoza@users.noreply.github.com >
2025-10-06 13:15:54 +05:30
ce90f9b2db
[FIX] Text to image training peft version ( #12434 )
...
Fix peft error
2025-10-06 08:24:54 +05:30
c3675d4c9b
[core] support QwenImage Edit Plus in modular ( #12416 )
...
* up
* up
* up
* up
* up
* up
* remove saves
* move things around a bit.
* get ready.
2025-10-05 21:57:13 +05:30
2b7deffe36
fix scale_shift_factor being on cpu for wan and ltx ( #12347 )
...
* wan fix scale_shift_factor being on cpu
* apply device cast to ltx transformer
* Apply style fixes
---------
Co-authored-by: Dhruv Nair <dhruv.nair@gmail.com >
Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
2025-10-05 09:23:38 +05:30
941ac9c3d9
[training-scripts] Make more examples UV-compatible (follow up on #12000 ) ( #12407 )
...
* make qwen and kontext uv compatible
* add torchvision
* add torchvision
* add datasets, bitsandbytes, prodigyopt
---------
Co-authored-by: Sayak Paul <spsayakpaul@gmail.com >
2025-10-03 17:46:47 +03:00
77c4e0932c
Merge branch 'main' into attn-refactor-blocks
2025-10-03 11:39:22 +05:30
fc322ed052
update
2025-10-03 11:38:16 +05:30
fed2c46482
update
2025-10-03 11:36:21 +05:30
66320f031a
update
2025-10-03 11:35:39 +05:30
7242b5ff62
FIX Test to ignore warning for enable_lora_hotswap ( #12421 )
...
I noticed that the test should be for the option check_compiled="ignore"
but it was using check_compiled="warn". This has been fixed, now the
correct argument is passed.
However, the fact that the test passed means that it was incorrect to
begin with. The way that logs are collected does not collect the
logger.warning call here (not sure why). To amend this, I'm now using
assertNoLogs. With this change, the test correctly fails when the wrong
argument is passed.
2025-10-02 20:57:11 +02:00
b4297967a0
[core] conditionally import torch distributed stuff. ( #12420 )
...
conditionally import torch distributed stuff.
2025-10-02 20:38:02 +05:30
9ae5b6299d
[ci] xfail failing tests in CI. ( #12418 )
...
xfail failing tests in CI.
2025-10-02 17:46:15 +05:30
814d710e56
[tests] cache non lora pipeline outputs. ( #12298 )
...
* cache non lora pipeline outputs.
* up
* up
* up
* up
* Revert "up"
This reverts commit 772c32e433cca03df7fc
2025-10-01 09:02:55 +05:30
cc5b31ffc9
[docs] Migrate syntax ( #12390 )
...
* change syntax
* make style
2025-09-30 10:11:19 -07:00
d7a1a0363f
[docs] CP ( #12331 )
...
* init
* feedback
* feedback
* feedback
* feedback
* feedback
* feedback
2025-09-30 09:33:41 -07:00
b59654544b
Install latest prerelease from huggingface_hub when installing transformers from main ( #12395 )
...
* Allow prerelease when installing transformers from main
* maybe better
* maybe better
* and now?
* just bored
* should be better
* works now
2025-09-30 17:02:33 +05:30
0e12ba7454
fix 3 xpu failures uts w/ latest pytorch ( #12408 )
...
fix xpu ut failures w/ latest pytorch
Signed-off-by: Yao, Matrix <matrix.yao@intel.com >
2025-09-30 14:07:48 +05:30
20fd00b14b
[Tests] Add single file tester mixin for Models and remove unittest dependency ( #12352 )
...
* update
* update
* update
* update
* update
2025-09-30 13:28:34 +05:30
76d4e416bc
[modular]some small fix ( #12307 )
...
* fix
* add mellon node registry
* style
* update docstring to include more info!
* support custom node mellon
* HTTPErrpr -> HfHubHTTPErrpr
* up
* Update src/diffusers/modular_pipelines/qwenimage/node_utils.py
2025-09-29 11:42:34 -10:00
c07fcf780a
[docs] Model formats ( #12256 )
...
* init
* config
* lora metadata
* feedback
* fix
* cache allocator warmup for from_single_file
* feedback
* feedback
2025-09-29 11:36:14 -07:00
ccedeca96e
[docs] Distributed inference ( #12285 )
...
* init
* feedback
---------
Co-authored-by: Sayak Paul <spsayakpaul@gmail.com >
2025-09-29 11:24:26 -07:00
64a5187d96
[quantization] feat: support aobaseconfig classes in TorchAOConfig ( #12275 )
...
* feat: support aobaseconfig classes.
* [docs] AOBaseConfig (#12302 )
init
Co-authored-by: Sayak Paul <spsayakpaul@gmail.com >
* up
* replace with is_torchao_version
* up
* up
---------
Co-authored-by: Steven Liu <59462357+stevhliu@users.noreply.github.com >
2025-09-29 18:04:18 +05:30
0a151115bb
Fix #12116 : preserve boolean dtype for attention masks in ChromaPipeline ( #12263 )
...
* fix: preserve boolean dtype for attention masks in ChromaPipeline
- Convert attention masks to bool and prevent dtype corruption
- Fix both positive and negative mask handling in _get_t5_prompt_embeds
- Remove float conversion in _prepare_attention_mask method
Fixes #12116
* test: add ChromaPipeline attention mask dtype tests
* test: add slow ChromaPipeline attention mask tests
* chore: removed comments
* refactor: removing redundant type conversion
* Remove dedicated dtype tests as per feedback
---------
Co-authored-by: Dhruv Nair <dhruv.nair@gmail.com >
2025-09-29 14:20:05 +05:30
19085ac8f4
Don't skip Qwen model tests for group offloading with disk ( #12382 )
...
u[
2025-09-29 13:08:05 +05:30
041501aea9
[docs] remove docstrings from repeated methods in lora_pipeline.py ( #12393 )
...
* start unbloating docstrings (save_lora_weights).
* load_lora_weights()
* lora_state_dict
* fuse_lora
* unfuse_lora
* load_lora_into_transformer
2025-09-26 22:38:43 +05:30
9c0944581a
[docs] slight edits to the attention backends docs. ( #12394 )
...
* slight edits to the attention backends docs.
* Update docs/source/en/optimization/attention_backends.md
Co-authored-by: Steven Liu <59462357+stevhliu@users.noreply.github.com >
---------
Co-authored-by: Steven Liu <59462357+stevhliu@users.noreply.github.com >
2025-09-26 21:50:16 +05:30
4588bbeb42
[CI] disable installing transformers from main in ci for now. ( #12397 )
...
* disable installing transformers from main in ci for now.
* up
* u[p
2025-09-26 18:41:17 +05:30
ec5449f3a1
Support both huggingface_hub v0.x and v1.x ( #12389 )
...
* Support huggingface_hub 0.x and 1.x
* httpx
2025-09-25 18:28:54 +02:00
310fdaf556
Introduce cache-dit to community optimization ( #12366 )
...
* docs: introduce cache-dit to diffusers
* docs: introduce cache-dit to diffusers
* docs: introduce cache-dit to diffusers
* docs: introduce cache-dit to diffusers
* docs: introduce cache-dit to diffusers
* docs: introduce cache-dit to diffusers
* docs: introduce cache-dit to diffusers
* misc: update examples link
* misc: update examples link
* docs: introduce cache-dit to diffusers
* docs: introduce cache-dit to diffusers
* docs: introduce cache-dit to diffusers
* docs: introduce cache-dit to diffusers
* docs: introduce cache-dit to diffusers
* Refine documentation for CacheDiT features
Updated the wording for clarity and consistency in the documentation. Adjusted sections on cache acceleration, automatic block adapter, patch functor, and hybrid cache configuration.
2025-09-24 10:50:57 -07:00
dcb6dd9b7a
Context Parallel w/ Ring & Ulysses & Unified Attention ( #11941 )
...
* update
* update
* add coauthor
Co-Authored-By: Dhruv Nair <dhruv.nair@gmail.com >
* improve test
* handle ip adapter params correctly
* fix chroma qkv fusion test
* fix fastercache implementation
* fix more tests
* fight more tests
* add back set_attention_backend
* update
* update
* make style
* make fix-copies
* make ip adapter processor compatible with attention dispatcher
* refactor chroma as well
* remove rmsnorm assert
* minify and deprecate npu/xla processors
* update
* refactor
* refactor; support flash attention 2 with cp
* fix
* support sage attention with cp
* make torch compile compatible
* update
* refactor
* update
* refactor
* refactor
* add ulysses backward
* try to make dreambooth script work; accelerator backward not playing well
* Revert "try to make dreambooth script work; accelerator backward not playing well"
This reverts commit 768d0ea6fadhruv.nair@gmail.com >
* apply review suggestions
* update docs
* add explanation
* make fix-copies
* add docstrings
* support passing parallel_config to from_pretrained
* apply review suggestions
* make style
* update
* Update docs/source/en/api/parallel.md
Co-authored-by: Aryan <aryan@huggingface.co >
* up
---------
Co-authored-by: Dhruv Nair <dhruv.nair@gmail.com >
Co-authored-by: sayakpaul <spsayakpaul@gmail.com >
2025-09-24 19:03:25 +05:30
043ab2520f
Fix WanVACEPipeline to allow prompt to be None and skip encoding step ( #12251 )
...
Fixed WanVACEPipeline to allow prompt to be None and skip encoding step
2025-09-24 15:15:04 +05:30
08c29020dd
fix marigold ut case fail on xpu ( #12350 )
...
Signed-off-by: Yao, Matrix <matrix.yao@intel.com >
2025-09-24 09:32:06 +05:30
7a58734994
xpu enabling for 4 cases ( #12345 )
...
Signed-off-by: Yao, Matrix <matrix.yao@intel.com >
2025-09-24 09:31:45 +05:30
9ef118509e
[tests] disable xformer tests for pipelines it isn't popular. ( #12277 )
...
disable xformer tests for pipelines it isn't popular.
2025-09-24 09:02:25 +05:30
7c54a7b38a
Fix Custom Code loading ( #12378 )
...
* update
* update
* update
2025-09-24 08:53:41 +05:30
09e777a3e1
[tests] Single scheduler in lora tests ( #12315 )
...
* single scheduler please.
* up
* up
* up
2025-09-24 08:36:50 +05:30
a72bc0c4bb
[docs] Attention backends ( #12320 )
...
* init
* feedback
* update
* feedback
* fixes
2025-09-23 10:59:46 -07:00
80de641c1c
Allow Automodel to support custom model code ( #12353 )
...
* update
* update
2025-09-23 07:31:42 -10:00
76810eca2b
[docs] Schedulers ( #12246 )
...
* init
* toctree
* scheduler suggestions
* toctree
2025-09-23 10:29:16 -07:00
1448b03585
[Fix] chroma docs ( #12360 )
...
* Fixes chroma docs
* fix docs
fixed docs are now consistent
2025-09-22 13:04:13 -07:00
DN6