feat: allow disk offload for diffuser models#3285

Merged

patrickvonplaten merged 85 commits intohuggingface:mainfrom

hari10599:hkr/disk-offload

May 22, 2023

Commits on Apr 29, 2023

Commits on May 20, 2023

add max_memory argument
hari10599
committed
Changed sample[0] to images[0] (#3304)

IliaLarchenko
authored and
hari10599
committed
Typo in tutorial (#3295)

IliaLarchenko
authored and
hari10599
committed
Torch compile graph fix (#3286)

patrickvonplaten
authored and
hari10599
committed
Postprocessing refactor img2img (#3268)

authored and
hari10599
committed
[Torch 2.0 compile] Fix more torch compile breaks (#3313)

authored and
hari10599
committed
fix: scale_lr and sync example readme and docs. (#3299)

sayakpaul
authored and
hari10599
committed
Update stable_diffusion.mdx (#3310)

mu94-csl
authored and
hari10599
committed
Fix missing variable assign in DeepFloyd-IF-II (#3315)

gitmylo
authored and
hari10599
committed
Correct doc build for patch releases (#3316)

patrickvonplaten
authored and
hari10599
committed
Add Stable Diffusion RePaint to community pipelines (#3320)

Markus-Pobitzer
authored and
hari10599
committed
Fix multistep dpmsolver for cosine schedule (suitable for deepfloyd-if) (#3314)

authored and
hari10599
committed
[docs] Improve LoRA docs (#3311)

stevhliu
authored and
hari10599
committed
Added input pretubation (#3292)

isamu-isozaki
authored and
hari10599
committed
Update write_own_pipeline.mdx (#3323)

csaybar
authored and
hari10599
committed
update controlling generation doc with latest goodies. (#3321)

sayakpaul
authored and
hari10599
committed
[Quality] Make style (#3341)

patrickvonplaten
authored and
hari10599
committed
Fix config dpm (#3343)

patrickvonplaten
authored and
hari10599
committed
Add the SDE variant of DPM-Solver and DPM-Solver++ (#3344)

LuChengTHU
authored and
hari10599
committed
Add upsample_size to AttnUpBlock2D, AttnDownBlock2D (#3275)

will-rice
authored and
hari10599
committed
Rename --only_save_embeds to --save_as_full_pipeline (#3206)

arrufat
authored and
hari10599
committed
[AudioLDM] Generalise conversion script (#3328)

authored and
hari10599
committed
Fix TypeError when using prompt_embeds and negative_prompt (#2982)

At-sushi
authored and
hari10599
committed
Fix pipeline class on README (#3345)

themrzmaster
authored and
hari10599
committed
Inpainting: typo in docs (#3331)

authored and
hari10599
committed
Add `use_Karras_sigmas` to LMSDiscreteScheduler (#3351)

Isotr0py
authored and
hari10599
committed
Batched load of textual inversions (#3277)

authored and
hari10599
committed
make fix-copies

patrickvonplaten
authored and
hari10599
committed
[docs] Fix docstring (#3334)

authored and
hari10599
committed
if dreambooth lora (#3360)

williamberman
authored and
hari10599
committed
Postprocessing refactor all others (#3337)

authored and
hari10599
committed
[docs] Improve safetensors docstring (#3368)

stevhliu
authored and
hari10599
committed
add: a warning message when using xformers in a PT 2.0 env. (#3365)

authored and
hari10599
committed
StableDiffusionInpaintingPipeline - resize image w.r.t height and width (#3322)

authored and
hari10599
committed
make style

patrickvonplaten
authored and
hari10599
committed
[docs] Adapt a model (#3326)

stevhliu
authored and
hari10599
committed
[docs] Load safetensors (#3333)

authored and
hari10599
committed
make style

patrickvonplaten
authored and
hari10599
committed
[Docs] Fix stable_diffusion.mdx typo (#3398)

sudowind
authored and
hari10599
committed
Support ControlNet v1.1 shuffle properly (#3340)

authored and
hari10599
committed
[Tests] better determinism (#3374)

sayakpaul
authored and
hari10599
committed
[docs] Add transformers to install (#3388)

stevhliu
authored and
hari10599
committed
[deepspeed] partial ZeRO-3 support (#3076)

authored and
hari10599
committed
Add omegaconf for tests (#3400)

patrickvonplaten
authored and
hari10599
committed
Fix various bugs with LoRA Dreambooth and Dreambooth script (#3353)

authored and
hari10599
committed
Fix docker file (#3402)

patrickvonplaten
authored and
hari10599
committed
fix: deepseepd_plugin retrieval from accelerate state (#3410)

sayakpaul
authored and
hari10599
committed
[Docs] Add `sigmoid` beta_scheduler to docstrings of relevant Schedulers (#3399)

authored and
hari10599
committed
Don't install accelerate and transformers from source (#3415)

patrickvonplaten
authored and
hari10599
committed
Don't install transformers and accelerate from source (#3414)

patrickvonplaten
authored and
hari10599
committed
Improve fast tests (#3416)

patrickvonplaten
authored and
hari10599
committed
attention refactor: the trilogy (#3387)

williamberman
authored and
hari10599
committed
[Docs] update the PT 2.0 optimization doc with latest findings (#3370)

authored and
hari10599
committed
Fix style rendering (#3433)

pcuenca
authored and
hari10599
committed
unCLIP scheduler do not use note (#3417)

williamberman
authored and
hari10599
committed
Replace deprecated command with environment file (#3409)

authored and
hari10599
committed
fix warning message pipeline loading (#3446)

patrickvonplaten
authored and
hari10599
committed
add stable diffusion tensorrt img2img pipeline (#3419)

asfiyab-nvidia
authored and
hari10599
committed
Refactor controlnet and add img2img and inpaint (#3386)

patrickvonplaten
authored and
hari10599
committed
[Scheduler] DPM-Solver (++) Inverse Scheduler (#3335)

authored and
hari10599
committed
[Docs] Fix incomplete docstring for resnet.py (#3438)

Laurent2916
authored and
hari10599
committed
fix tiled vae blend extent range (#3384)

superlabs-dev
authored and
hari10599
committed
Small update to "Next steps" section (#3443)

pcuenca
authored and
hari10599
committed
Allow arbitrary aspect ratio in IFSuperResolutionPipeline (#3298)

authored and
hari10599
committed
Adding 'strength' parameter to StableDiffusionInpaintingPipeline (#3424)

authored and
hari10599
committed
[WIP] Bugfix - Pipeline.from_pretrained is broken when the pipeline is partially downloaded (#3448)

vimarshc
authored and
hari10599
committed
Fix gradient checkpointing bugs in freezing part of models (requires_grad=False) (#3404)

authored and
hari10599
committed
Make dreambooth lora more robust to orig unet (#3462)

patrickvonplaten
authored and
hari10599
committed
Reduce peak VRAM by releasing large attention tensors (as soon as they're unnecessary) (#3463)

cmdr2
authored and
hari10599
committed
Add min snr to text2img lora training script (#3459)

wfng92
authored and
hari10599
committed
Add inpaint lora scale support (#3460)

Glaceon-Hyy
authored and
hari10599
committed
[From ckpt] Fix from_ckpt (#3466)

patrickvonplaten
authored and
hari10599
committed
Update full dreambooth script to work with IF (#3425)

williamberman
authored and
hari10599
committed
Add IF dreambooth docs (#3470)

williamberman
authored and
hari10599
committed
parameterize pass single args through tuple (#3477)

williamberman
authored and
hari10599
committed
attend and excite tests disable determinism on the class level (#3478)

williamberman
authored and
hari10599
committed
dreambooth docs torch.compile note (#3471)

authored and
hari10599
committed
add: if entry in the dreambooth training docs. (#3472)

sayakpaul
authored and
hari10599
committed
[docs] Textual inversion inference (#3473)

authored and
hari10599
committed
[docs] Distributed inference (#3376)

stevhliu
authored and
hari10599
committed
[{Up,Down}sample1d] explicit view kernel size as number elements in flattened indices (#3479)

williamberman
authored and
hari10599
committed
mps & onnx tests rework (#3449)

pcuenca
authored and
hari10599
committed
Merge branch 'huggingface:main' into hkr/disk-offload
hari10599
authored