@@ -447,7 +447,7 @@ def get_platform():
447447 'system' : platform .system (),
448448 'release' : release ,
449449 'python' : platform .python_version (),
450- 'docker' : os .environ .get ('SD_INSTALL_DEBUG ' , None ) is not None ,
450+ 'docker' : os .environ .get ('SD_DOCKER ' , None ) is not None ,
451451 # 'host': platform.node(),
452452 # 'version': platform.version(),
453453 }
@@ -492,7 +492,7 @@ def check_diffusers():
492492 t_start = time .time ()
493493 if args .skip_all or args .skip_git :
494494 return
495- sha = 'b785ddb654e4be3ae0066e231734754bdb2a191c ' # diffusers commit hash
495+ sha = '7b100ce589b917d4c116c9e61a6ec46d4f2ab062 ' # diffusers commit hash
496496 pkg = pkg_resources .working_set .by_key .get ('diffusers' , None )
497497 minor = int (pkg .version .split ('.' )[1 ] if pkg is not None else 0 )
498498 cur = opts .get ('diffusers_version' , '' ) if minor > 0 else ''
@@ -625,6 +625,9 @@ def install_rocm_zluda():
625625 else :
626626 torch_command = os .environ .get ('TORCH_COMMAND' , f'torch torchvision --index-url https://download.pytorch.org/whl/rocm{ rocm .version } )
627627
628+ if os .environ .get ('TRITON_COMMAND' , None ) is None :
629+ os .environ .setdefault ('TRITON_COMMAND' , 'skip' ) # pytorch auto installs pytorch-triton-rocm as a dependency instead
630+
628631 if sys .version_info < (3 , 11 ):
629632 ort_version = os .environ .get ('ONNXRUNTIME_VERSION' , None )
630633 if rocm .version is None or float (rocm .version ) > 6.0 :
@@ -659,22 +662,39 @@ def install_rocm_zluda():
659662
660663def install_ipex (torch_command ):
661664 t_start = time .time ()
662- check_python (supported_minors = [10 ,11 ], reason = 'IPEX backend requires Python 3.10 or 3.11' )
665+ # Python 3.12 will cause compatibility issues with other dependencies
666+ # IPEX supports Python 3.12 so don't block it but don't advertise it in the error message
667+ check_python (supported_minors = [9 , 10 , 11 , 12 ], reason = 'IPEX backend requires Python 3.9, 3.10 or 3.11' )
663668 args .use_ipex = True # pylint: disable=attribute-defined-outside-init
664669 log .info ('IPEX: Intel OneAPI toolkit detected' )
670+
665671 if os .environ .get ("NEOReadDebugKeys" , None ) is None :
666672 os .environ .setdefault ('NEOReadDebugKeys' , '1' )
667673 if os .environ .get ("ClDeviceGlobalMemSizeAvailablePercent" , None ) is None :
668674 os .environ .setdefault ('ClDeviceGlobalMemSizeAvailablePercent' , '100' )
675+ if os .environ .get ("SYCL_CACHE_PERSISTENT" , None ) is None :
676+ os .environ .setdefault ('SYCL_CACHE_PERSISTENT' , '1' ) # Jit cache
677+
669678 if os .environ .get ("PYTORCH_ENABLE_XPU_FALLBACK" , None ) is None :
670- os .environ .setdefault ('PYTORCH_ENABLE_XPU_FALLBACK' , '1' )
679+ os .environ .setdefault ('PYTORCH_ENABLE_XPU_FALLBACK' , '1' ) # CPU fallback for unsupported ops
680+ if os .environ .get ("OverrideDefaultFP64Settings" , None ) is None :
681+ os .environ .setdefault ('OverrideDefaultFP64Settings' , '1' )
682+ if os .environ .get ("IGC_EnableDPEmulation" , None ) is None :
683+ os .environ .setdefault ('IGC_EnableDPEmulation' , '1' ) # FP64 Emulation
684+ if os .environ .get ('IPEX_FORCE_ATTENTION_SLICE' , None ) is None :
685+ # XPU PyTorch doesn't support Flash Atten or Memory Atten yet so Battlemage goes OOM without this
686+ os .environ .setdefault ('IPEX_FORCE_ATTENTION_SLICE' , '1' )
687+
671688 if "linux" in sys .platform :
672- torch_command = os .environ .get ('TORCH_COMMAND' , 'torch==2.5.1+cxx11.abi torchvision==0.20.1+cxx11.abi intel-extension-for-pytorch==2.5.10+xpu oneccl_bind_pt==2.5.0+xpu --extra-index-url https://pytorch-extension.intel.com/release-whl/stable/xpu/cn/' )
673- # torch_command = os.environ.get('TORCH_COMMAND', 'torch torchvision --index-url https://download.pytorch.org/whl/test/xpu') # test wheels are stable previews, significantly slower than IPEX
674- # os.environ.setdefault('TENSORFLOW_PACKAGE', 'tensorflow==2.15.1 intel-extension-for-tensorflow[xpu]==2.15.0.1')
689+ # default to US server. If The China server is needed, change .../release-whl/stable/xpu/us/ to .../release-whl/stable/xpu/cn/
690+ torch_command = os .environ .get ('TORCH_COMMAND' , 'torch==2.5.1+cxx11.abi torchvision==0.20.1+cxx11.abi intel-extension-for-pytorch==2.5.10+xpu oneccl_bind_pt==2.5.0+xpu --extra-index-url https://pytorch-extension.intel.com/release-whl/stable/xpu/us/' )
691+ if os .environ .get ('TRITON_COMMAND' , None ) is None :
692+ os .environ .setdefault ('TRITON_COMMAND' , '--pre pytorch-triton-xpu==3.1.0+91b14bf559 --index-url https://download.pytorch.org/whl/nightly/xpu' )
693+ # os.environ.setdefault('TENSORFLOW_PACKAGE', 'tensorflow==2.15.1 intel-extension-for-tensorflow[xpu]==2.15.0.2')
675694 else :
676- torch_command = os .environ .get ('TORCH_COMMAND' , '--pre torch torchvision --index-url https://download.pytorch.org/whl/nightly/xpu' ) # torchvision doesn't exist on test/stable branch for windows
677- install (os .environ .get ('OPENVINO_PACKAGE' , 'openvino==2024.5.0' ), 'openvino' , ignore = True )
695+ torch_command = os .environ .get ('TORCH_COMMAND' , 'torch==2.6.0+xpu torchvision==0.21.0+xpu --index-url https://download.pytorch.org/whl/test/xpu' )
696+
697+ install (os .environ .get ('OPENVINO_PACKAGE' , 'openvino==2024.6.0' ), 'openvino' , ignore = True )
678698 install ('nncf==2.7.0' , ignore = True , no_deps = True ) # requires older pandas
679699 install (os .environ .get ('ONNXRUNTIME_PACKAGE' , 'onnxruntime-openvino' ), 'onnxruntime-openvino' , ignore = True )
680700 ts ('ipex' , t_start )
@@ -683,6 +703,8 @@ def install_ipex(torch_command):
683703
684704def install_openvino (torch_command ):
685705 t_start = time .time ()
706+ # Python 3.12 will cause compatibility issues with other dependencies.
707+ # OpenVINO supports Python 3.12 so don't block it but don't advertise it in the error message
686708 check_python (supported_minors = [9 , 10 , 11 , 12 ], reason = 'OpenVINO backend requires Python 3.9, 3.10 or 3.11' )
687709 log .info ('OpenVINO: selected' )
688710 if sys .platform == 'darwin' :
@@ -726,11 +748,22 @@ def install_torch_addons():
726748 install ('optimum-quanto==0.2.6' , 'optimum-quanto' )
727749 if not args .experimental :
728750 uninstall ('wandb' , quiet = True )
729- if triton_command is not None :
751+ if triton_command is not None and triton_command != 'skip' :
730752 install (triton_command , 'triton' , quiet = True )
731753 ts ('addons' , t_start )
732754
733755
756+ # check cudnn
757+ def check_cudnn ():
758+ import site
759+ site_packages = site .getsitepackages ()
760+ cuda_path = os .environ .get ('CUDA_PATH' , '' )
761+ for site_package in site_packages :
762+ folder = os .path .join (site_package , 'nvidia' , 'cudnn' , 'lib' )
763+ if os .path .exists (folder ) and folder not in cuda_path :
764+ os .environ ['CUDA_PATH' ] = f"{ cuda_path } { folder }
765+
766+
734767# check torch version
735768def check_torch ():
736769 t_start = time .time ()
@@ -842,6 +875,7 @@ def check_torch():
842875 return
843876 if not args .skip_all :
844877 install_torch_addons ()
878+ check_cudnn ()
845879 if args .profile :
846880 pr .disable ()
847881 print_profile (pr , 'Torch' )
@@ -1056,7 +1090,7 @@ def install_optional():
10561090 install ('gfpgan' )
10571091 install ('clean-fid' )
10581092 install ('pillow-jxl-plugin==1.3.1' , ignore = True )
1059- install ('optimum-quanto=0.2.6' , ignore = True )
1093+ install ('optimum-quanto== 0.2.6' , ignore = True )
10601094 install ('bitsandbytes==0.45.0' , ignore = True )
10611095 install ('pynvml' , ignore = True )
10621096 install ('ultralytics==8.3.40' , ignore = True )
0 commit comments