Machine Learning Engineer
-
Hugging Face
- Paris, France
- https://linkedin.com/in/anton-lozhkov/
- @anton_lozhkov
Block or Report
Block or report anton-l
Report abuse
Contact GitHub support about this user’s behavior. Learn more about reporting abuse.
Report abusePinned
-
huggingface/diffusers Public
🤗 Diffusers: State-of-the-art diffusion models for image and audio generation in PyTorch -
huggingface/transformers Public
🤗 Transformers: State-of-the-art Machine Learning for Pytorch, TensorFlow, and JAX.
457 contributions in the last year
Less
More
Contribution activity
August 2022
Created 26 commits in 2 repositories
Created 2 repositories
- anton-l/diffusers-ci Python
- anton-l/stable-diffusion Jupyter Notebook
Created a pull request in huggingface/diffusers that received 5 comments
Support one-string prompts and custom image size in LDM
Fixes #211
Makes sure that LDMTextToImagePipeline can be called with pipe("some prompt"), similar to StableDiffusionPipeline
+22
−12
•
5
comments
Opened 13 other pull requests in 3 repositories
huggingface/diffusers
7
merged
1
open
-
Style the
scriptsdirectory - Pin black==22.3 to keep a stable --preview flag
- [WIP] Add RePaint inpainting pipeline
-
Add
is_torch_available,is_flax_available -
Revive
make quality - Add K-LMS scheduler from k-diffusion
- Add an alternative Karras et al. stochastic scheduler for VE models
- Support training with a local image folder
anton-l/diffusers-ci
3
open
huggingface/notebooks
2
merged
Reviewed 24 pull requests in 1 repository
huggingface/diffusers
24 pull requests
- Reproducible images by supplying latents to pipeline
- Add training options
- added test workflow and fixed failing test
- split tests_modeling_utils
- Add image2image example script.
- [Loading] allow modules to be loaded in fp16
- add add_noise method in LMSDiscreteScheduler, PNDMScheduler
- Pipeline to device
- Add safety module
- Allow passing non-default modules to pipeline
- Add K-LMS scheduler from k-diffusion
- [PNDM] Stable diffusion
- [Half precision] Make sure half-precision is correct
- allow custom height, width in StableDiffusionPipeline
- add scaled_linear schedule in PNDM and DDPM
- allow pndm scheduler to be used with ldm pipeline
- add attention up/down blocks for VAE
- Add an alternative Karras et al. stochastic scheduler for VE models
- [LDMTextToImagePipeline] make text model generic
-
Fix arg key for
dataset_nameincreate_model_card - [VAE] fix the downsample block in Encoder.
- [UNet2DConditionModel] add cross_attention_dim as an argument
- Add issue templates for feature requests and bug reports
- Allow DDPM scheduler to use model's predicated variance
1
contribution
in private repositories
Aug 4





