GitLab

* refac: docstrings in training_utils.py

* fix: manual edits

* run make style

* add docstring at cast_training_params

---------
Co-authored-by: Sayak Paul <spsayakpaul@gmail.com>

* fix: retain memory utility.

* fix

* quality

* free_memory.

to avoid "FutureWarning: transformers.deepspeed module is deprecated and will be removed in a future version. Please import deepspeed modules directly from transformers.integrations"
Co-authored-by: Sayak Paul <spsayakpaul@gmail.com>

Add extra performance features for EMAModel, torch._foreach operations and better support for non-blocking CPU offloading (#7685)

* Add support for _foreach operations and non-blocking to EMAModel

* default foreach to false

* add non-blocking EMA offloading to SD1.5 T2I example script

* fix whitespace

* move foreach to cli argument

* linting

* Update README.md re: EMA weight training

* correct args.foreach_ema

* add tests for foreach ema

* code quality

* add foreach to from_pretrained

* default foreach false

* fix linting

---------
Co-authored-by: Sayak Paul <spsayakpaul@gmail.com>
Co-authored-by: drhead <a@a.a>

refactor the density and weighting utilities.

Co-authored-by: Jimmy <39@🇺🇸

.com>
Co-authored-by: Sayak Paul <spsayakpaul@gmail.com>
Co-authored-by: YiYi Xu <yixu310@gmail.com>

A new function compute_dream_and_update_latents has been added to the
training utilities that allows you to do DREAM rectified training in line
with the paper https://arxiv.org/abs/2312.00210. The method can be used
with an extra argument in the train_text_to_image.py script.

Co-authored-by: Jimmy <39@🇺🇸.com>

* Add npu support

* fix for code quality check

* fix for code quality check

* add

* remove transformer

---------
Co-authored-by: yiyixuxu <yixu310@gmail,com>

create a utility for casting the lora params during training.

* fix: training resume from fp16.

* add: comment

* remove residue from another branch.

* remove more residues.

* thanks to Younes; no hacks.

* style.

* clean things a bit and modularize _set_state_dict_into_text_encoder

* add comment about the fix detailed.

* Make WDS pipeline interpolation type configurable.

* Make the VAE encoding batch size configurable.

* Make lora_alpha and lora_dropout configurable for LCM LoRA scripts.

* Generalize scalings_for_boundary_conditions function and make the timestep scaling configurable.

* Make LoRA target modules configurable for LCM-LoRA scripts.

* Move resolve_interpolation_mode to src/diffusers/training_utils.py and make interpolation type configurable in non-WDS script.

* apply suggestions from review

* fix: duplicate unet prefix problem.

* Update src/diffusers/loaders/lora.py
Co-authored-by: Patrick von Platen <patrick.v.platen@gmail.com>

---------
Co-authored-by: Patrick von Platen <patrick.v.platen@gmail.com>

* use loralinear instead of depecrecated lora attn procs.

* fix parameters()

* fix saving

* add back support for add kv proj.

* fix: param accumul,ation.

* propagate the changes.

add compute_snr() to training utils.

* up

* fix more

* Apply suggestions from code review

* fix more

* fix more

* Check it

* Remove 16:8

* fix more

* fix more

* fix more

* up

* up

* Test only stable diffusion

* Test only two files

* up

* Try out spinning up processes that can be killed

* up

* Apply suggestions from code review

* up

* up

* [deepspeed] partial ZeRO-3 support

* cleanup

* improve deepspeed fixes

* Improve

* make style

---------
Co-authored-by: Patrick von Platen <patrick.v.platen@gmail.com>

* add store and restore() methods to EMAModel.

* Update src/diffusers/training_utils.py
Co-authored-by: Patrick von Platen <patrick.v.platen@gmail.com>

* make style with doc builder

* remove explicit listing.

* Apply suggestions from code review
Co-authored-by: Will Berman <wlbberman@gmail.com>

* Apply suggestions from code review
Co-authored-by: Patrick von Platen <patrick.v.platen@gmail.com>

* chore: better variable naming.

* better treatment of temp_stored_params
Co-authored-by: patil-suraj <surajp815@gmail.com>

* make style

* remove temporary params from earth 🌎



* make fix-copies.

---------
Co-authored-by: Patrick von Platen <patrick.v.platen@gmail.com>
Co-authored-by: Will Berman <wlbberman@gmail.com>
Co-authored-by: patil-suraj <surajp815@gmail.com>

* EMA: fix `state_dict()` & add `cur_decay_value`

* EMA: fix a bug in `load_state_dict()`

'float' object (`state_dict["power"]`) has no attribute 'get'.

* del train_unconditional_ort.py

* better accelerated saving

* up

* finish

* finish

* uP

* up

* up

* fix

* Apply suggestions from code review

* correct ema

* Remove @

* up

* Apply suggestions from code review
Co-authored-by: Pedro Cuenca <pedro@huggingface.co>

* Update docs/source/en/training/dreambooth.mdx
Co-authored-by: Pedro Cuenca <pedro@huggingface.co>

---------
Co-authored-by: Pedro Cuenca <pedro@huggingface.co>

Possible typo introduced in 7c82a16f

* improve EMA

* style

* one EMA model

* quality

* fix tests

* fix test

* Apply suggestions from code review
Co-authored-by: Pedro Cuenca <pedro@huggingface.co>

* re organise the unconditional script

* backwards compatibility

* default to init values for some args

* fix ort script

* issubclass => isinstance

* update state_dict

* docstr

* doc

* Apply suggestions from code review
Co-authored-by: Pedro Cuenca <pedro@huggingface.co>

* use .to if device is passed

* deprecate device

* make flake happy

* fix typo
Co-authored-by: patil-suraj <surajp815@gmail.com>
Co-authored-by: Pedro Cuenca <pedro@huggingface.co>
Co-authored-by: Patrick von Platen <patrick.v.platen@gmail.com>

* add a doc page for each pipeline under api/pipelines/stable_diffusion
* add pipeline examples to docstrings
* updated stable_diffusion_2 page
* updated default markdown syntax to list methods based on https://github.com/huggingface/diffusers/pull/1870


* add function decorator
Co-authored-by: yiyixuxu <yixu@Yis-MacBook-Pro.lan>
Co-authored-by: Patrick von Platen <patrick.v.platen@gmail.com>

* [Vae and AutoencoderKL clean]

* save intermediate finished work

* more progress

* more progress

* finish modeling code

* save intermediate

* finish

* Correct tests

* Add torch_device to the VE pipeline

* Mark the training test with slow

* Catch unused params in DDP

* Fix proj_out, add test

ema model stepping done automatically now