patrickvonplaten
Follow
Highlights
- Pro
Pinned
2,757 contributions in the last year
We're celebrating 100 million developers!
Contribution activity
January 2023
Created 70 commits in 6 repositories
Created 2 repositories
Created a pull request in huggingface/diffusers that received 108 comments
[LoRA] Add LoRA training script
Update: Training seems to work fine -> see some results here (after 4min of training on a A100): https://wandb.ai/patrickvonplaten/stable_diffusion…
+2,288
−664
•
108
comments
Opened 40 other pull requests in 6 repositories
huggingface/diffusers
3
open
31
merged
1
closed
- [Design philosopy] Create official doc
- Allow lora from pipeline
- [LoRA] Make sure LoRA can be disabled after it's run
-
Don't call the Hub if
local_files_onlyis specifiied - [Import Utils] Fix naming
-
[Bump version] 0.13.0dev0 & Deprecate
predict_epsilon - Improve transformers versions handling
- Allow directly passing text embeddings to Stable Diffusion Pipeline for prompt weighting
- [Paint by example] Fix cpu offload for paint by example
- [From pretrained] Don't download .safetensors files if safetensors is…
- Correct Pix2Pix example
-
[WIP] Use
acceleratesave & loading hooks to have better checkpoint structure - [WIP] Add nice contribution doc
- [Save Pretrained] Remove dead code lines that can accidentally remove pytorch files
- [Dit] Fix dit tests
- [Lora] Model card
- [Docs] No more autocast
- [Black] Update black library
- [Community] Fix merger
- [Conversion] Improve safetensors
- Fix slow tests
- Example tests
- [CPU offload] correct cpu offload
- [WIP] Don't merge / CPU offload example
- [Community] Correct checkpoint merger
- Some pull requests not shown.
huggingface/transformers
1
open
huggingface/speechbox
1
merged
intel/neural-compressor
1
open
timothybrooks/instruct-pix2pix
1
closed
huggingface/accelerate
1
merged
Reviewed 85 pull requests in 7 repositories
huggingface/diffusers
25 pull requests
- Section on using LoRA alpha / scale
- Fix typo
- [Enhancement] Improve status and error reporting
-
Adding
use_safetensorsargument to give more control to users - [lora] Allow finetuning text encoder with lora
-
Typo fix:
torwards->towards - Fix typos in loaders.py
- remove redundant allow_patterns
- Allow lora from pipeline
-
Adding some
safetensorsdocs. - Fix model card of LoRA
- Fix unable to save_pretrained when using pathlib
- [Bug]: fix DDPM scheduler arbitrary infer steps count.
- For legacy inpaint pipeline, scale mask tensor to latent size and avoid noise added to final latents
- Add dynamic thresholding to DDIM and DDPM
- make scaling factor a config arg of vae/vqvae
- fuse attention mask
- [textual inversion] Allow validation images
- Reproducibility 3/3
- Allow directly passing text embeddings to Stable Diffusion Pipeline for prompt weighting
- Remove wandb from text_to_image requirements.txt
- [docs] [dreambooth] note random crop
- Correct Pix2Pix example
- add load textual inversion embeddings to stable diffusion
- [Feat] checkpoint_merger works on local models as well as ones that use safetensors
- Some pull request reviews not shown.
huggingface/transformers
4 pull requests
huggingface/speechbox
1 pull request
huggingface/accelerate
1 pull request
huggingface/blog
1 pull request
timothybrooks/instruct-pix2pix
1 pull request
huggingface/huggingface_hub
1 pull request
Created an issue in huggingface/accelerate that received 9 comments
[Discussion] save_state / load_state don't use custom saving/loading methods.
save_state currently serializes models always as a simple pytorch_model.bin file, but no additional metadata is saved such as the model config.
Cou…
9
comments