Public/automatic
1
0
Fork 0
mirror of https://github.com/vladmandic/automatic synced 2026-09-19 09:14:35 +02:00
Code Issues Packages Projects Releases Wiki Activity

IPEX 4GB alloc detection and log driver version

Browse Source
This commit is contained in:
Disty0
2025-01-22 18:15:25 +03:00
parent 9f3a34d0a3
commit af35296a68
5 changed files with 26 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files
modules/intel/ipex/diffusers.py
+2 -3
View File
@@ -6,7 +6,6 @@ from diffusers.models.attention_processor import Attention
# pylint: disable=protected-access, missing-function-docstring, line-too-long
device_supports_fp64 = torch.xpu.has_fp64_dtype() if hasattr(torch.xpu, "has_fp64_dtype") else torch.xpu.get_device_properties("xpu").has_fp64
attention_slice_rate = float(os.environ.get('IPEX_ATTENTION_SLICE_RATE', 4))
@@ -342,10 +341,10 @@ class AttnProcessor:
return hidden_states
def ipex_diffusers():
def ipex_diffusers(device_supports_fp64=False, can_allocate_plus_4gb=False):
diffusers.utils.torch_utils.fourier_filter = fourier_filter
#ARC GPUs can't allocate more than 4GB to a single block:
if os.environ.get('IPEX_FORCE_ATTENTION_SLICE', '0') == '1' or (not device_supports_fp64 and os.environ.get('IPEX_FORCE_ATTENTION_SLICE', '0') == '0'):
if not can_allocate_plus_4gb:
diffusers.models.attention_processor.SlicedAttnProcessor = SlicedAttnProcessor
diffusers.models.attention_processor.AttnProcessor = AttnProcessor
if not device_supports_fp64: