aboutsummaryrefslogtreecommitdiffstats
path: root/modules/devices.py
AgeCommit message (Collapse)AuthorLines
2023-08-09split shared.py into multiple files; should resolve all circular reference ↵AUTOMATIC1111-9/+1
import errors related to shared.py
2023-08-09rework RNG to use generators instead of generating noises beforehandAUTOMATIC1111-79/+2
2023-08-03rework torchsde._brownian.brownian_interval replacement to use ↵AUTOMATIC1111-6/+38
device.randn_local and respect the NV setting.
2023-08-03add NV option for Random number generator source setting, which allows to ↵AUTOMATIC1111-2/+37
generate same pictures on CPU/AMD/Mac as on NVidia videocards.
2023-07-11Fix MPS cache cleanupAarni Koskela-2/+3
Importing torch does not import torch.mps so the call failed.
2023-07-08added torch.mps.empty_cache() to torch_gc()AUTOMATIC1111-0/+3
changed a bunch of places that use torch.cuda.empty_cache() to use torch_gc() instead
2023-06-05Remove a bunch of unused/vestigial codeAarni Koskela-7/+0
As found by Vulture and some eyes
2023-05-21run basic torch calculation at startup in parallel to reduce the performance ↵AUTOMATIC-0/+18
impact of first generation
2023-05-10ruff auto fixesAUTOMATIC-1/+1
2023-04-29rename CPU RNG to RNG source in settings, add infotext and parameters ↵AUTOMATIC-2/+2
copypaste support to RNG source
2023-04-18Option to use CPU for random number generation.Deciare-2/+6
Makes a given manual seed generate the same images across different platforms, independently of the GPU architecture in use. Fixes #9613.
2023-02-01Refactor Mac specific code to a separate filebrkirch-45/+7
Move most Mac related code to a separate file, don't even load it unless web UI is run under macOS.
2023-02-01Refactor MPS fixes to CondFuncbrkirch-36/+14
2023-02-01MPS fix is still needed :(brkirch-0/+3
Apparently I did not test with large enough images to trigger the bug with torch.narrow on MPS
2023-01-28Merge pull request #7309 from brkirch/fix-embeddingsAUTOMATIC1111-3/+8
Fix embeddings, upscalers, and refactor `--upcast-sampling`
2023-01-28Remove MPS fix no longer needed for PyTorchbrkirch-3/+0
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.
2023-01-28Refactor conditional casting, fix upscalersbrkirch-0/+8
2023-01-27clarify the option to disable NaN check.AUTOMATIC-0/+2
2023-01-27remove the need to place configs near modelsAUTOMATIC-4/+8
2023-01-25Add UI setting for upcasting attention to float32brkirch-1/+5
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.
2023-01-25Add option for float32 sampling with float16 UNetbrkirch-0/+2
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().
2023-01-19Merge pull request #6922 from brkirch/cumsum-fixAUTOMATIC1111-4/+7
Improve cumsum fix for MPS
2023-01-17Fix cumsum for MPS in newer torchbrkirch-4/+7
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.
2023-01-17disable the new NaN check for the CIAUTOMATIC-0/+3
2023-01-16Add a check and explanation for tensor with all NaNs.AUTOMATIC-0/+28
2023-01-05Add support for PyTorch nightly and local buildsbrkirch-5/+23
2022-12-17Add numpy fix for MPS on PyTorch 1.12.1brkirch-0/+9
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.
2022-12-03add built-in extension systemAUTOMATIC-1/+10
add support for adding upscalers in extensions move LDSR, ScuNET and SwinIR to built-in extensions
2022-12-03add comment for #4407 and remove seemingly unnecessary cudnn.enabledAUTOMATIC-1/+3
2022-12-03fix #4407 breaking UI entirely for card other than ones related to the PRAUTOMATIC-4/+2
2022-12-03Merge pull request #4407 from yoinked-h/patch-1AUTOMATIC1111-0/+7
Fix issue with 16xx cards
2022-11-30Rework MPS randn fix, add randn_like fixbrkirch-12/+3
torch.manual_seed() already sets a CPU generator, so there is no reason to create a CPU generator manually. torch.randn_like also needs a MPS fix for k-diffusion, but a torch hijack with randn_like already exists so it can also be used for that.
2022-11-27Merge pull request #4918 from brkirch/pytorch-fixesAUTOMATIC1111-7/+24
Fixes for PyTorch 1.12.1 when using MPS
2022-11-27eliminate duplicated code from #5095AUTOMATIC-19/+11
2022-11-26torch.cuda.empty_cache() defaults to cuda:0 device unless explicitly set ↵Matthew McGoogan-2/+12
otherwise first. Updating torch_gc() to use the device set by --device-id if specified to avoid OOM edge cases on multi-GPU systems.
2022-11-21Add fixes for PyTorch 1.12.1brkirch-1/+27
Fix typo "MasOS" -> "macOS" If MPS is available and PyTorch is an earlier version than 1.13: * Monkey patch torch.Tensor.to to ensure all tensors sent to MPS are contiguous * Monkey patch torch.nn.functional.layer_norm to ensure input tensor is contiguous (required for this program to work with MPS on unmodified PyTorch 1.12.1)
2022-11-17Revert "MPS Upscalers Fix"brkirch-9/+0
This reverts commit 768b95394a8500da639b947508f78296524f1836.
2022-11-12change formatting to match the main program in devices.pyAUTOMATIC-5/+16
2022-11-12Update devices.py源文雨-1/+1
2022-11-12Fix wrong mps selection below MasOS 12.3源文雨-3/+10
2022-11-08actual better fixpepe10-gpu-5/+2
thanks C43H66N12O12S2
2022-11-07terrible hackpepe10-gpu-2/+9
2022-11-0616xx card fixpepe10-gpu-0/+3
cudnn
2022-10-25MPS Upscalers Fixbrkirch-0/+4
Get ESRGAN, SCUNet, and SwinIR working correctly on MPS by ensuring memory is contiguous for tensor views before sending to MPS device.
2022-10-25Remove BSRGAN from --use-cpu, add SwinIRbrkirch-1/+1
2022-10-22remove parsing command line from devices.pyAUTOMATIC-9/+5
2022-10-22implement CUDA device selection by IDExtraltodeus-3/+18
2022-10-14Add 'interrogate' and 'all' choices to --use-cpubrkirch-1/+1
* Add 'interrogate' and 'all' choices to --use-cpu * Change type for --use-cpu argument to str.lower, so that choices are case insensitive
2022-10-10--no-half-vaeAUTOMATIC-1/+5
2022-10-04Merge branch 'master' into cpu-cmdline-optbrkirch-0/+10
generated by cgit v1.2.3 (git 2.39.1) at 2025-10-23 04:56:06 +0000