Age | Commit message (Collapse) | Author | |
---|---|---|---|
2024-01-09 | Apply correct inference precision implementation | Kohaku-Blueleaf | |
2024-01-09 | linting and debugs | Kohaku-Blueleaf | |
2024-01-09 | Fix bugs when arg dtype doesn't match | KohakuBlueleaf | |
2024-01-09 | improve efficiency and support more device | Kohaku-Blueleaf | |
2023-12-31 | change import statements for #14478 | AUTOMATIC1111 | |
2023-12-31 | Add utility to inspect a model's parameters (to get dtype/device) | Aarni Koskela | |
2023-12-03 | Merge branch 'dev' into test-fp8 | Kohaku-Blueleaf | |
2023-12-02 | Merge pull request #14171 from Nuullll/ipex | AUTOMATIC1111 | |
Initial IPEX support for Intel Arc GPU | |||
2023-12-02 | Merge branch 'dev' into test-fp8 | Kohaku-Blueleaf | |
2023-12-02 | Merge pull request #14131 from read-0nly/patch-1 | AUTOMATIC1111 | |
Update devices.py - Make 'use-cpu all' actually apply to 'all' | |||
2023-12-02 | Disable ipex autocast due to its bad perf | Nuullll | |
2023-11-30 | Initial IPEX support | Nuullll | |
2023-11-27 | Update devices.py | obsol | |
fixes issue where "--use-cpu" all properly makes SD run on CPU but leaves ControlNet (and other extensions, I presume) pointed at GPU, causing a crash in ControlNet caused by a mismatch between devices between SD and CN https://github.com/AUTOMATIC1111/stable-diffusion-webui/issues/14097 | |||
2023-11-19 | Better naming | Kohaku-Blueleaf | |
2023-11-19 | Use options instead of cmd_args | Kohaku-Blueleaf | |
2023-10-28 | Add MPS manual cast | KohakuBlueleaf | |
2023-10-28 | ManualCast for 10/16 series gpu | Kohaku-Blueleaf | |
2023-10-24 | Add CPU fp8 support | Kohaku-Blueleaf | |
Since norm layer need fp32, I only convert the linear operation layer(conv2d/linear) And TE have some pytorch function not support bf16 amp in CPU. I add a condition to indicate if the autocast is for unet. | |||
2023-09-09 | fix for crash when running #12924 without --device-id | AUTOMATIC1111 | |
2023-08-31 | More accurate check for enabling cuDNN benchmark on 16XX cards | catboxanon | |
2023-08-09 | split shared.py into multiple files; should resolve all circular reference ↵ | AUTOMATIC1111 | |
import errors related to shared.py | |||
2023-08-09 | rework RNG to use generators instead of generating noises beforehand | AUTOMATIC1111 | |
2023-08-03 | rework torchsde._brownian.brownian_interval replacement to use ↵ | AUTOMATIC1111 | |
device.randn_local and respect the NV setting. | |||
2023-08-03 | add NV option for Random number generator source setting, which allows to ↵ | AUTOMATIC1111 | |
generate same pictures on CPU/AMD/Mac as on NVidia videocards. | |||
2023-07-11 | Fix MPS cache cleanup | Aarni Koskela | |
Importing torch does not import torch.mps so the call failed. | |||
2023-07-08 | added torch.mps.empty_cache() to torch_gc() | AUTOMATIC1111 | |
changed a bunch of places that use torch.cuda.empty_cache() to use torch_gc() instead | |||
2023-06-05 | Remove a bunch of unused/vestigial code | Aarni Koskela | |
As found by Vulture and some eyes | |||
2023-05-21 | run basic torch calculation at startup in parallel to reduce the performance ↵ | AUTOMATIC | |
impact of first generation | |||
2023-05-10 | ruff auto fixes | AUTOMATIC | |
2023-04-29 | rename CPU RNG to RNG source in settings, add infotext and parameters ↵ | AUTOMATIC | |
copypaste support to RNG source | |||
2023-04-18 | Option to use CPU for random number generation. | Deciare | |
Makes a given manual seed generate the same images across different platforms, independently of the GPU architecture in use. Fixes #9613. | |||
2023-02-01 | Refactor Mac specific code to a separate file | brkirch | |
Move most Mac related code to a separate file, don't even load it unless web UI is run under macOS. | |||
2023-02-01 | Refactor MPS fixes to CondFunc | brkirch | |
2023-02-01 | MPS fix is still needed :( | brkirch | |
Apparently I did not test with large enough images to trigger the bug with torch.narrow on MPS | |||
2023-01-28 | Merge pull request #7309 from brkirch/fix-embeddings | AUTOMATIC1111 | |
Fix embeddings, upscalers, and refactor `--upcast-sampling` | |||
2023-01-28 | Remove MPS fix no longer needed for PyTorch | brkirch | |
The torch.narrow fix was required for nightly PyTorch builds for a while to prevent a hard crash, but newer nightly builds don't have this issue. | |||
2023-01-28 | Refactor conditional casting, fix upscalers | brkirch | |
2023-01-27 | clarify the option to disable NaN check. | AUTOMATIC | |
2023-01-27 | remove the need to place configs near models | AUTOMATIC | |
2023-01-25 | Add UI setting for upcasting attention to float32 | brkirch | |
Adds "Upcast cross attention layer to float32" option in Stable Diffusion settings. This allows for generating images using SD 2.1 models without --no-half or xFormers. In order to make upcasting cross attention layer optimizations possible it is necessary to indent several sections of code in sd_hijack_optimizations.py so that a context manager can be used to disable autocast. Also, even though Stable Diffusion (and Diffusers) only upcast q and k, unfortunately my findings were that most of the cross attention layer optimizations could not function unless v is upcast also. | |||
2023-01-25 | Add option for float32 sampling with float16 UNet | brkirch | |
This also handles type casting so that ROCm and MPS torch devices work correctly without --no-half. One cast is required for deepbooru in deepbooru_model.py, some explicit casting is required for img2img and inpainting. depth_model can't be converted to float16 or it won't work correctly on some systems (it's known to have issues on MPS) so in sd_models.py model.depth_model is removed for model.half(). | |||
2023-01-19 | Merge pull request #6922 from brkirch/cumsum-fix | AUTOMATIC1111 | |
Improve cumsum fix for MPS | |||
2023-01-17 | Fix cumsum for MPS in newer torch | brkirch | |
The prior fix assumed that testing int16 was enough to determine if a fix is needed, but a recent fix for cumsum has int16 working but not bool. | |||
2023-01-17 | disable the new NaN check for the CI | AUTOMATIC | |
2023-01-16 | Add a check and explanation for tensor with all NaNs. | AUTOMATIC | |
2023-01-05 | Add support for PyTorch nightly and local builds | brkirch | |
2022-12-17 | Add numpy fix for MPS on PyTorch 1.12.1 | brkirch | |
When saving training results with torch.save(), an exception is thrown: "RuntimeError: Can't call numpy() on Tensor that requires grad. Use tensor.detach().numpy() instead." So for MPS, check if Tensor.requires_grad and detach() if necessary. | |||
2022-12-03 | add built-in extension system | AUTOMATIC | |
add support for adding upscalers in extensions move LDSR, ScuNET and SwinIR to built-in extensions | |||
2022-12-03 | add comment for #4407 and remove seemingly unnecessary cudnn.enabled | AUTOMATIC | |
2022-12-03 | fix #4407 breaking UI entirely for card other than ones related to the PR | AUTOMATIC | |