3527 Commits

This Branch

This Branch

All Branches

Author	SHA1	Message	Date
cmdr2	14e45bfd24	Early support for AMD 9070 (and Navi 4x series)	2025-05-30 13:11:24 +05:30
cmdr2	46c4aca7b7	Recognize 5060 Ti and a few other recent GPUs	2025-05-30 10:58:36 +05:30
cmdr2	38eb8f934e	Support more GPUs (workstation and mining); Fix torchruntime test	2025-04-29 10:40:24 +05:30
cmdr2	0393c907d7	Fix torchruntime to not force cu128 on older NVIDIA GPUs (which don't support it)	2025-04-29 10:08:42 +05:30
cmdr2	f9c664dc3c	Tell the user to reinstall ED if they're trying to use a 50xx GPU with Python 3.8-based Easy Diffusion	2025-04-27 21:29:14 +05:30
cmdr2	59ca8b0336	Fix FileNotFoundError when installing torch on Windows PCs without powershell in their system PATH variable	2025-04-25 20:45:37 +05:30
cmdr2	e0a4d43087	Use pytorch 2.7 (with cuda 12.8) on new installations with NVIDIA gpus	2025-04-24 15:51:15 +05:30
cmdr2	a33737b991	Auto-upgrade torch for NVIDIA 50xx series. Fix for #1918	2025-04-23 15:23:54 +05:30
cmdr2	bac23290dd	Fix for incorrect upgrade logic for torchruntime. Affects the workaround in #1918	2025-04-23 15:23:49 +05:30
cmdr2	8380ba97c5	Update README.md	2025-04-02 10:13:42 +05:30
cmdr2	45c6865b2d	Fix for blackwell GPUs on new installations	2025-03-29 14:35:55 +05:30
cmdr2	1b3d048bf2	Use safetensors as the default model instead of ckpt	2025-03-11 10:25:13 +05:30
cmdr2	a675161e47	Fix nvidia 50xx support	2025-03-10 06:21:18 +05:30
cmdr2	3a9f71d17a	Fix compatibility with python 3.9 and directml	2025-03-07 10:39:58 +05:30
cmdr2	b09be681e6	Check for half-precision on GeForce MX450	2025-03-07 10:28:28 +05:30
cmdr2	4330434835	Fix Granite Ridge APU device type	2025-03-06 12:48:46 +05:30
cmdr2	3836b91ae1	Recognize some more cards (5070, Granite Ridge) via a torchruntime upgrade v3.0.9c	2025-03-06 12:07:42 +05:30
cmdr2	72c4e47619	Support Blackwell (NVIDIA 5060/5070/5080/5090)	2025-03-04 16:23:43 +05:30
cmdr2	afae421cee	Use python 3.9 by default	2025-03-04 15:28:38 +05:30
cmdr2	0d0ec4ee56	sdkit 2.0.22.7/2.0.15.16 - python 3.9 compatibility	2025-03-04 15:10:17 +05:30
cmdr2	55a31c77e6	torchruntime upgrade - fix bug with multiple GPUs of the same model	2025-03-04 11:28:03 +05:30
cmdr2	43d2642b68	Update README.md	2025-02-21 17:08:30 +05:30
cmdr2	9dc2154027	sdkit upgrade - fixes loras with numpy arrays	2025-02-20 10:12:35 +05:30
cmdr2	fd49ba5dbc	Update README.md	2025-02-18 11:59:33 +05:30
cmdr2	3e71054150	Potential fix for #1902	2025-02-18 11:14:42 +05:30
cmdr2	8d6c0de262	Recognize Phoenix3 and Phoenix4 AMD APUs	2025-02-18 11:14:37 +05:30
cmdr2	561fe0cc79	Use torchruntime for installing torch/torchvision on the users' PC, instead of the custom logic used here. torchruntime was built from the custom logic used here, and covers a lot more scenarios and supports more devices	2025-02-13 11:56:24 +05:30
cmdr2	26cbc30407	Use the correct size of the image when used as the input. Code credit: @AvidGameFan	2025-02-13 11:33:29 +05:30
cmdr2	7a1e2c4190	Hotfix for missing torchruntime on new installations	2025-02-10 19:31:26 +05:30
cmdr2	7b0a17a3ab	Temporary fix for #1899	2025-02-10 18:15:31 +05:30
cmdr2	302426f5d4	Another fix for mps backend	2025-02-10 10:05:35 +05:30
cmdr2	9dc9ea3825	Fix broken mps backend	2025-02-10 09:55:31 +05:30
cmdr2	2a24a49f6b	torchruntime 1.9.2 - fixes a bug on cpu platforms	2025-02-08 16:15:12 +05:30
cmdr2	5e5e39c285	changelog	2025-02-08 15:10:50 +05:30
cmdr2	cd8365558a	Remove hardcoded references to torch.cuda; Use torchruntime and sdkit's device utilities instead	2025-02-08 15:08:46 +05:30
cmdr2	2e4623736a	sdkit 2.0.22.3 or 2.0.15.12 - fixes a regression on mac	2025-02-03 21:32:06 +05:30
cmdr2	7f3a4383c7	sdkit 2.0.22.2 or 2.0.15.11 - install torchruntime	2025-02-03 21:32:02 +05:30
cmdr2	6d6b528aad	sdkit 2.0.22 or 2.0.15.9	2025-02-03 10:58:47 +05:30
cmdr2	76485ab1e7	Move the half precision bug check logic to sdkit	2025-02-03 10:58:43 +05:30
cmdr2	68d67248f4	[sdkit update] v2.0.22 (and v2.0.15.8 for LTS)	2025-02-03 10:58:39 +05:30
cmdr2	81119a5893	Skip sdkit/diffusers install if it's in developer mode	2025-01-28 09:57:04 +05:30
cmdr2	554559f5ce	changelog	2025-01-28 09:56:16 +05:30
cmdr2	b3cc415359	Temporarily remove torch 2.5 from the list, since it doesn't work with Python 3.8. More on this in future commits	2025-01-28 09:55:09 +05:30
cmdr2	5ac44de6c7	Even older rocm versions	2025-01-09 16:39:44 +05:30
cmdr2	a7a78a40d0	Allow older rocm versions	2025-01-09 16:00:18 +05:30
cmdr2	fea24cee90	Update README.md	2025-01-07 11:29:20 +05:30
cmdr2	20d77a85a1	Upgrade the version of torch used for rocm for Navi 30+, and point to the broader torch URL	2025-01-07 10:32:25 +05:30
cmdr2	0687e7b020	Update the index url for AMD ROCm torch install	2025-01-06 19:49:06 +05:30
cmdr2	75e4dc25dc	Extend the list of supported torch, CUDA and ROCm versions	2025-01-06 19:32:10 +05:30
cmdr2	7e635caec8	version bump for wmic deprecation	2025-01-04 18:12:49 +05:30