Ecosyste.ms: Issues

An open API service for providing issue and pull request metadata for open source projects.

GitHub / crowsonkb/k-diffusion issues and pull requests

#114 - Use `torch.special.expm1`

Pull Request - State: open - Opened by kit1980 30 days ago

#113 - Why zero initialize patch_out in hourglass transformer?

Issue - State: open - Opened by Luciennnnnnn 3 months ago - 1 comment

#111 - Scaled cosine similarity can cause NaNs

Issue - State: open - Opened by bobcao3 6 months ago

#110 - Use of DCT in loss function?

Issue - State: open - Opened by sundeco 7 months ago

#109 - remove append_dims from to_d to allow sigmas on cpu

Pull Request - State: open - Opened by drhead 9 months ago

#107 - Patch size 2 for HDiT

Issue - State: open - Opened by yuanzhi-zhu 9 months ago - 1 comment

#106 - bad sampler name: DPM++ 2M Karras

Issue - State: open - Opened by itsklimov 9 months ago

#105 - How to inference the trained model?

Issue - State: open - Opened by sudheendra-krishnamurthy 9 months ago

#104 - fix: DPM-Solver++(2m & 3M) SDE.

Pull Request - State: open - Opened by davertor 9 months ago

#103 - Confirm JSON config for FFHQ-1024?

Issue - State: open - Opened by tin-sely 10 months ago - 2 comments

#99 - sample.py: Error while deserializing header: HeaderTooLarge

Issue - State: closed - Opened by drscotthawley 12 months ago - 4 comments

#98 - Questions about soft-min-snr loss

Issue - State: open - Opened by hmicrobe 12 months ago - 1 comment

#96 - R2 setup in the paper

Issue - State: closed - Opened by nicolas-dufour almost 1 year ago - 4 comments

#95 - Scaling from 128x128, to 256x256, 512x512 and 1024x1024?

Issue - State: open - Opened by tin-sely about 1 year ago - 4 comments

#94 - Consider Making Normalization Optional in MappingFeedForwardBlock

Issue - State: open - Opened by mnslarcher about 1 year ago - 2 comments

#92 - Pretrained models of Hourglass Diffusion Transformers

Issue - State: open - Opened by fenglinglwb about 1 year ago

#91 - `--mixed-precision` doesnt work with img transformer 2

Issue - State: closed - Opened by yoinked-h about 1 year ago - 2 comments

#90 - [draft] Noise Perturbation from Ning et al 2022

Pull Request - State: closed - Opened by zaptrem about 1 year ago

#89 - How to cite this?

Issue - State: closed - Opened by exx8 about 1 year ago - 1 comment

#87 - [Feature request]Support DDIM-Solver and Seniorious

Issue - State: open - Opened by sdbds about 1 year ago

#86 - fix "UnboundLocalError: local variable 'h' referenced before assignment"

Pull Request - State: closed - Opened by lewisay over 1 year ago - 2 comments

#85 - Stabilizing DPM++2M SDE for SDXL

Issue - State: open - Opened by LuChengTHU over 1 year ago - 2 comments

#83 - Drop rotary-embedding-torch dependency

Pull Request - State: closed - Opened by akx over 1 year ago - 1 comment

#82 - Use torchsde-brownian instead of full torchsde

Pull Request - State: closed - Opened by akx over 1 year ago

#80 - Fixed UnboundLocalError

Pull Request - State: open - Opened by ka1tte over 1 year ago - 1 comment

#79 - Loglikelihood Evaluation Question

Issue - State: open - Opened by Me-lab over 1 year ago - 14 comments

#78 - fix mistakes in calculation FID: sqrt root matrix, eps 1e-6, equation

Pull Request - State: closed - Opened by shonenkov over 1 year ago - 4 comments

#77 - bugfix: mps device precision error

Pull Request - State: closed - Opened by afpro over 1 year ago - 1 comment

#76 - LMS sampler question

Issue - State: closed - Opened by catboxanon over 1 year ago - 2 comments

#74 - fix: local variable 'h' referenced before assignment

Pull Request - State: closed - Opened by storyicon over 1 year ago

#73 - Eliminate jsonmerge dependency

Pull Request - State: open - Opened by pharmapsychotic over 1 year ago - 2 comments

#72 - UnboundLocalError: local variable 'h' referenced before assignment

Issue - State: open - Opened by Yaiol over 1 year ago - 2 comments

#70 - Conditional *image* generation (img2img)

Issue - State: open - Opened by CallShaul over 1 year ago - 2 comments

#69 - Move h_last assignment to avoid cases where h is undefined

Pull Request - State: closed - Opened by ArrowM over 1 year ago - 1 comment

#68 - Add transform_last for BrownianTreeNoiseSampler

Pull Request - State: open - Opened by Disty0 over 1 year ago

#67 - added a citation.cff

Pull Request - State: open - Opened by kashif over 1 year ago

#65 - Add missing check for empty tensors

Pull Request - State: open - Opened by ilikenwf over 1 year ago

#64 - Common Diffusion Noise Schedules and Sample Steps are Flawed

Issue - State: open - Opened by sdbds over 1 year ago - 1 comment

#63 - time consuming

Issue - State: open - Opened by fuchao01 over 1 year ago

#62 - Fixing "SeedSequence expects int or sequence of ints for entropy not N.0"

Pull Request - State: open - Opened by chmatse over 1 year ago - 1 comment

#61 - Fix lint issues using ruff

Pull Request - State: open - Opened by akx over 1 year ago - 2 comments

#60 - Move training-only dependencies to [train] extra

Pull Request - State: open - Opened by akx over 1 year ago - 3 comments

#59 - Feature Request: Support the "correct" SDE variant of DPM-Solver++

Issue - State: closed - Opened by LuChengTHU almost 2 years ago - 5 comments

#58 - Addition of UniPC?

Issue - State: open - Opened by ImpossibleAd436 almost 2 years ago

#57 - Accelerate is not working fine

Issue - State: open - Opened by frangam almost 2 years ago

#56 - sample_dpmpp_2m has a bug?

Issue - State: open - Opened by hallatore almost 2 years ago - 19 comments

#54 - ModuleNotFoundError: No module named 'torch_directml'

Issue - State: closed - Opened by CloudJoe almost 2 years ago

#52 - Both the FID scores and visualization are pretty bad in FFHQ dataset

Issue - State: open - Opened by IrohCao-NYU about 2 years ago - 1 comment

#51 - FID scores from paper

Issue - State: open - Opened by darius-lam about 2 years ago - 4 comments

#50 - Learning rate scheduler broken

Issue - State: closed - Opened by kyleliang919 about 2 years ago - 2 comments

#49 - Sample many files more efficiently

Issue - State: open - Opened by hotohoto about 2 years ago - 1 comment

#47 - Add an example on how to use k-diffusion with `diffusers`

Pull Request - State: closed - Opened by patrickvonplaten about 2 years ago - 2 comments

#46 - Noise samplers

Pull Request - State: closed - Opened by crowsonkb about 2 years ago

#45 - (Missing) Loss weights for the Diffusion Loss?

Issue - State: closed - Opened by mbreuss about 2 years ago

#44 - Reproducibility work on the sampling code

Pull Request - State: closed - Opened by ardacihaner about 2 years ago - 2 comments

#43 - Stabilize the sampling of DPM-Solver++2M by a stabilizing trick

Issue - State: open - Opened by LuChengTHU over 2 years ago - 22 comments

#41 - Simple evaluation script

Pull Request - State: open - Opened by pcuenca over 2 years ago

#40 - Add the newest DPM-Solver

Issue - State: closed - Opened by LuChengTHU over 2 years ago - 6 comments

#39 - Do not merge! Credits for DEIS

Pull Request - State: open - Opened by qsh-zh over 2 years ago

#38 - remove `clean-fid`

Pull Request - State: closed - Opened by space-pope over 2 years ago - 2 comments

#37 - sample sigma scheduler bug

Issue - State: open - Opened by lockmatrix over 2 years ago - 2 comments

#36 - Question: Is it possible to predict step count of DPM adaptive?

Issue - State: closed - Opened by C43H66N12O12S2 over 2 years ago - 1 comment

#35 - added support for non square Images

Pull Request - State: open - Opened by p-sodmann over 2 years ago

#33 - main -> master in gh action

Pull Request - State: closed - Opened by rom1504 over 2 years ago

#32 - Add github action to automatically push to pypi on Release x.y.z commit

Pull Request - State: closed - Opened by rom1504 over 2 years ago - 1 comment

#28 - Constant starting noise for demo() images

Pull Request - State: open - Opened by bob80333 over 2 years ago

#27 - Req: TQDM trange() update frequency / miniters option

Issue - State: open - Opened by torridgristle over 2 years ago

#26 - Allow installing via pip install -e <repo link> - Adds a setup.py

Pull Request - State: closed - Opened by cmdr2 over 2 years ago - 4 comments

#24 - [stale]

Issue - State: open - Opened by ozanciga over 2 years ago

#23 - Implement time step discretization for Karras samplers

Pull Request - State: open - Opened by Birch-san over 2 years ago - 16 comments

#22 - 'Refactored by Sourcery' (#1)

Pull Request - State: closed - Opened by AdamOswald over 2 years ago

#21 - Loading kdiffy.py results in this error

Issue - State: open - Opened by TheArtOfTheVoid over 2 years ago

#20 - How CLIP guided sampling works

Issue - State: open - Opened by htzheng over 2 years ago - 1 comment

#19 - Cache FID computation

Issue - State: open - Opened by neverix over 2 years ago

#18 - Manual random seed

Pull Request - State: closed - Opened by neverix over 2 years ago - 1 comment

#17 - Image inpainting

Issue - State: open - Opened by neverix over 2 years ago

#16 - Fine-tuning/compatible pre-trained models

Issue - State: open - Opened by kjhenner over 2 years ago - 2 comments

#15 - Example how to use the VDenoiser/OpenAIDenoiser/CompVisDenoiser wrappers?

Issue - State: open - Opened by Quasimondo over 2 years ago - 5 comments

#14 - Example for how to use the grow option?

Issue - State: open - Opened by Quasimondo over 2 years ago - 14 comments

#13 - CUBLAS_STATUS_ALLOC_FAILED error

Issue - State: open - Opened by kjhenner over 2 years ago - 1 comment

#12 - add gradient accumulation support

Pull Request - State: closed - Opened by tmabraham over 2 years ago - 1 comment

#11 - HuggingFace Datasets support with new config

Pull Request - State: closed - Opened by tmabraham over 2 years ago - 1 comment

#10 - About the data normalization and 'sigma_data'.

Issue - State: closed - Opened by jwliu-cc over 2 years ago - 2 comments

#9 - Config for training other resolutions

Issue - State: closed - Opened by eliahuhorwitz over 2 years ago - 4 comments

#8 - Use more standard fid/kid calculation

Pull Request - State: closed - Opened by samedii over 2 years ago - 18 comments