Public/automatic
1
0
Fork 0
mirror of https://github.com/vladmandic/automatic synced 2026-09-19 17:24:32 +02:00
Code Issues Packages Projects Releases Wiki Activity

IPEX Torch 2.10 support

Browse Source
This commit is contained in:
Disty0
2025-09-02 16:15:33 +03:00
parent 83a11c5839
commit 87a29642d3
3 changed files with 37 additions and 30 deletions
Download Patch File Download Diff File Expand all files Collapse all files
modules/intel/ipex/__init__.py
+24 -22
View File
@@ -8,7 +8,11 @@ except Exception:
has_ipex = False
from .hijacks import ipex_hijacks
torch_version = float(torch.__version__[:3])
torch_version = torch.__version__[:4]
if torch_version[-1] not in {"0", "1", "2", "3", "4", "5", "6", "7", "8", "9"}:
torch_version = torch_version[:-1]
torch_version = torch_version.split(".")
torch_version[0], torch_version[1] = int(torch_version[0]), int(torch_version[1])
# pylint: disable=protected-access, missing-function-docstring, line-too-long
@@ -20,7 +24,7 @@ def ipex_init(): # pylint: disable=too-many-statements
try:
# force xpu device on torch compile and triton
# import inductor utils to get around lazy import
from torch._inductor import utils as torch_inductor_utils # pylint: disable=import-error, unused-import # noqa: F401
from torch._inductor import utils as torch_inductor_utils # pylint: disable=import-error, unused-import # noqa: F401,RUF100
torch._inductor.utils.GPU_TYPES = ["xpu"]
torch._inductor.utils.get_gpu_type = lambda *args, **kwargs: "xpu"
from triton import backends as triton_backends # pylint: disable=import-error
@@ -71,7 +75,7 @@ def ipex_init(): # pylint: disable=too-many-statements
torch.cuda.__file__ = torch.xpu.__file__
# torch.cuda.is_current_stream_capturing = torch.xpu.is_current_stream_capturing
if torch_version < 2.3:
if torch_version[0] < 2 or (torch_version[0] == 2 and torch_version[1] < 3):
torch.cuda._initialization_lock = torch.xpu.lazy_init._initialization_lock
torch.cuda._initialized = torch.xpu.lazy_init._initialized
torch.cuda._is_in_bad_fork = torch.xpu.lazy_init._is_in_bad_fork
@@ -104,6 +108,12 @@ def ipex_init(): # pylint: disable=too-many-statements
torch.cuda.BoolStorage = torch.xpu.BoolStorage
torch.cuda.ComplexFloatStorage = torch.xpu.ComplexFloatStorage
torch.cuda.ComplexDoubleStorage = torch.xpu.ComplexDoubleStorage
torch._C._cuda_getCurrentRawStream = ipex._C._getCurrentRawStream
ipex._C._DeviceProperties.multi_processor_count = ipex._C._DeviceProperties.gpu_subslice_count
ipex._C._DeviceProperties.major = 12
ipex._C._DeviceProperties.minor = 1
ipex._C._DeviceProperties.L2_cache_size = 16*1024*1024 # A770 and A750
else:
torch.cuda._initialization_lock = torch.xpu._initialization_lock
torch.cuda._initialized = torch.xpu._initialized
@@ -114,23 +124,29 @@ def ipex_init(): # pylint: disable=too-many-statements
torch.cuda.threading = torch.xpu.threading
torch.cuda.traceback = torch.xpu.traceback
if torch_version < 2.5:
torch._C._cuda_getCurrentRawStream = torch._C._xpu_getCurrentRawStream
torch._C._XpuDeviceProperties.multi_processor_count = torch._C._XpuDeviceProperties.gpu_subslice_count
torch._C._XpuDeviceProperties.major = 12
torch._C._XpuDeviceProperties.minor = 1
torch._C._XpuDeviceProperties.L2_cache_size = 16*1024*1024 # A770 and A750
if torch_version[0] < 2 or (torch_version[0] == 2 and torch_version[1] < 5):
torch.cuda.os = torch.xpu.os
torch.cuda.Device = torch.xpu.Device
torch.cuda.warnings = torch.xpu.warnings
torch.cuda.classproperty = torch.xpu.classproperty
torch.UntypedStorage.cuda = torch.UntypedStorage.xpu
if torch_version < 2.7:
if torch_version[0] < 2 or (torch_version[0] == 2 and torch_version[1] < 7):
torch.cuda.Tuple = torch.xpu.Tuple
torch.cuda.List = torch.xpu.List
if torch_version < 2.8:
if torch_version[0] < 2 or (torch_version[0] == 2 and torch_version[1] < 8):
if has_ipex:
torch.cuda.memory_summary = torch.xpu.memory_summary
torch.cuda.memory_snapshot = torch.xpu.memory_snapshot
if torch_version < 2.9:
if torch_version[0] < 2 or (torch_version[0] == 2 and torch_version[1] < 9):
# torch._int_mm via onednn quantized matmul is supported with torch 2.9
# ipex 2.7+ has the same torch._int_mm support as torch 2.9 but doesn't support torch.compile
# torch._int_mm directly uses onednn quantized matmul
@@ -145,7 +161,7 @@ def ipex_init(): # pylint: disable=too-many-statements
# torch.compile fix
from .int_mm import qlinear_unary
torch._inductor.mkldnn_lowerings.register_onednn_fusion_ops.qlinear_unary = qlinear_unary
except Exception as e:
except Exception:
pass
# Memory:
@@ -178,20 +194,6 @@ def ipex_init(): # pylint: disable=too-many-statements
torch.cuda.seed_all = torch.xpu.seed_all
torch.cuda.initial_seed = torch.xpu.initial_seed
# C
if torch_version < 2.3:
torch._C._cuda_getCurrentRawStream = ipex._C._getCurrentRawStream
ipex._C._DeviceProperties.multi_processor_count = ipex._C._DeviceProperties.gpu_subslice_count
ipex._C._DeviceProperties.major = 12
ipex._C._DeviceProperties.minor = 1
ipex._C._DeviceProperties.L2_cache_size = 16*1024*1024 # A770 and A750
else:
torch._C._cuda_getCurrentRawStream = torch._C._xpu_getCurrentRawStream
torch._C._XpuDeviceProperties.multi_processor_count = torch._C._XpuDeviceProperties.gpu_subslice_count
torch._C._XpuDeviceProperties.major = 12
torch._C._XpuDeviceProperties.minor = 1
torch._C._XpuDeviceProperties.L2_cache_size = 16*1024*1024 # A770 and A750
# Fix functions with ipex:
# torch.xpu.mem_get_info always returns the total memory as free memory
torch.xpu.mem_get_info = lambda device=None: [(torch.xpu.get_device_properties(device).total_memory - torch.xpu.memory_reserved(device)), torch.xpu.get_device_properties(device).total_memory]
modules/intel/ipex/diffusers.py
+4 -4
View File
@@ -1,7 +1,7 @@
from functools import wraps
import torch
import diffusers # pylint: disable=import-error
from diffusers.utils import torch_utils # pylint: disable=import-error, unused-import # noqa: F401
from diffusers.utils import torch_utils # pylint: disable=import-error, unused-import # noqa: F401,RUF100
# pylint: disable=protected-access, missing-function-docstring, line-too-long
@@ -114,9 +114,9 @@ def ipex_diffusers(device_supports_fp64=False):
diffusers.utils.torch_utils.fourier_filter = fourier_filter
if not device_supports_fp64:
# get around lazy imports
from diffusers.models import embeddings as diffusers_embeddings # pylint: disable=import-error, unused-import # noqa: F401
from diffusers.models import transformers as diffusers_transformers # pylint: disable=import-error, unused-import # noqa: F401
from diffusers.models import controlnets as diffusers_controlnets # pylint: disable=import-error, unused-import # noqa: F401
from diffusers.models import embeddings as diffusers_embeddings # pylint: disable=import-error, unused-import # noqa: F401,RUF100
from diffusers.models import transformers as diffusers_transformers # pylint: disable=import-error, unused-import # noqa: F401,RUF100
from diffusers.models import controlnets as diffusers_controlnets # pylint: disable=import-error, unused-import # noqa: F401,RUF100
diffusers.models.embeddings.get_1d_sincos_pos_embed_from_grid = get_1d_sincos_pos_embed_from_grid
diffusers.models.embeddings.FluxPosEmbed = FluxPosEmbed
diffusers.models.embeddings.apply_rotary_emb = apply_rotary_emb
modules/intel/ipex/hijacks.py
+9 -4
View File
@@ -6,11 +6,16 @@ import numpy as np
from modules import devices, errors
torch_version = float(torch.__version__[:3])
torch_version = torch.__version__[:4]
if torch_version[-1] not in {"0", "1", "2", "3", "4", "5", "6", "7", "8", "9"}:
torch_version = torch_version[:-1]
torch_version = torch_version.split(".")
torch_version[0], torch_version[1] = int(torch_version[0]), int(torch_version[1])
device_supports_fp64 = torch.xpu.has_fp64_dtype() if hasattr(torch.xpu, "has_fp64_dtype") else torch.xpu.get_device_properties(devices.device).has_fp64
if os.environ.get('IPEX_FORCE_ATTENTION_SLICE', '0') == '0':
if torch_version >= 2.7:
if torch_version[0] > 2 or (torch_version[0] == 2 and torch_version[1] >= 7):
use_dynamic_attention = False # torch 2.7 has flash atten support
elif (torch.xpu.get_device_properties(devices.device).total_memory / 1024 / 1024 / 1024) > 4.1:
try:
@@ -289,7 +294,7 @@ def UntypedStorage_init(*args, device=None, **kwargs):
else:
return original_UntypedStorage_init(*args, device=device, **kwargs)
if torch_version >= 2.4:
if torch_version[0] > 2 or (torch_version[0] == 2 and torch_version[1] >= 4):
original_UntypedStorage_to = torch.UntypedStorage.to
@wraps(torch.UntypedStorage.to)
def UntypedStorage_to(self, *args, device=None, **kwargs):
@@ -407,7 +412,7 @@ class torch_Generator(original_torch_Generator):
# Hijack Functions:
def ipex_hijacks():
global device_supports_fp64
if torch_version >= 2.4:
if torch_version[0] > 2 or (torch_version[0] == 2 and torch_version[1] >= 4):
torch.UntypedStorage.cuda = UntypedStorage_cuda
torch.UntypedStorage.to = UntypedStorage_to
torch.tensor = torch_tensor