Public/automatic
1
0
Fork 0
mirror of https://github.com/vladmandic/automatic synced 2026-09-20 01:31:13 +02:00
Code Issues Packages Projects Releases Wiki Activity

ROCm and Zluda don't fallback to CPU and clenup strings

Browse Source
This commit is contained in:
Disty0
2025-09-27 11:32:46 +03:00
parent 579b1f3175
commit 71fde8a897
6 changed files with 26 additions and 45 deletions
Download Patch File Download Diff File Expand all files Collapse all files
modules/intel/ipex/__init__.py
+1 -1
View File
@@ -165,7 +165,7 @@ def ipex_init(): # pylint: disable=too-many-statements
pass
# Memory:
if 'linux' in sys.platform and "WSL2" in os.popen("uname -a").read():
if "linux" in sys.platform and "WSL2" in os.popen("uname -a").read():
torch.xpu.empty_cache = lambda: None
torch.cuda.empty_cache = torch.xpu.empty_cache
modules/intel/ipex/attention.py
+2 -2
View File
@@ -8,8 +8,8 @@ from functools import cache, wraps
# ARC GPUs can't allocate more than 4GB to a single block so we slice the attention layers
dynamic_attention_slice_rate = float(os.environ.get('IPEX_SDPA_SLICE_TRIGGER_RATE', 1))
dynamic_attention_trigger_rate = float(os.environ.get('IPEX_ATTENTION_SLICE_RATE', 0.5))
dynamic_attention_slice_rate = float(os.environ.get("IPEX_SDPA_SLICE_TRIGGER_RATE", "1"))
dynamic_attention_trigger_rate = float(os.environ.get("IPEX_ATTENTION_SLICE_RATE", "0.5"))
# Find something divisible with the input_tokens
@cache
modules/intel/ipex/hijacks.py
+6 -6
View File
@@ -80,8 +80,8 @@ def torch_get_autocast_dtype(device_type=None):
# IPEX 2.5 and above has partial support but doesn't really work most of the time.
original_interpolate = torch.nn.functional.interpolate
@wraps(torch.nn.functional.interpolate)
def interpolate(tensor, size=None, scale_factor=None, mode='nearest', align_corners=None, recompute_scale_factor=None, antialias=False): # pylint: disable=too-many-arguments
if mode in {'bicubic', 'bilinear'}:
def interpolate(tensor, size=None, scale_factor=None, mode="nearest", align_corners=None, recompute_scale_factor=None, antialias=False): # pylint: disable=too-many-arguments
if mode in {"bicubic", "bilinear"}:
return_device = tensor.device
return_dtype = tensor.dtype
return original_interpolate(tensor.to("cpu", dtype=torch.float32), size=size, scale_factor=scale_factor, mode=mode,
@@ -94,8 +94,8 @@ def interpolate(tensor, size=None, scale_factor=None, mode='nearest', align_corn
# SwinIR BF16:
original_functional_pad = torch.nn.functional.pad
@wraps(torch.nn.functional.pad)
def functional_pad(input, pad, mode='constant', value=None):
if mode == 'reflect' and input.dtype == torch.bfloat16:
def functional_pad(input, pad, mode="constant", value=None):
if mode == "reflect" and input.dtype == torch.bfloat16:
return original_functional_pad(input.to(torch.float32), pad, mode=mode, value=value).to(dtype=torch.bfloat16)
else:
return original_functional_pad(input, pad, mode=mode, value=value)
@@ -365,13 +365,13 @@ def ipex_hijacks():
except Exception:
pass
if os.environ.get('IPEX_FORCE_ATTENTION_SLICE', '0') == '0':
if os.environ.get("IPEX_FORCE_ATTENTION_SLICE", "0") == "0":
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
else:
use_dynamic_attention = True
else:
use_dynamic_attention = bool(os.environ.get('IPEX_FORCE_ATTENTION_SLICE', '0') == '1')
use_dynamic_attention = bool(os.environ.get("IPEX_FORCE_ATTENTION_SLICE", "0") == "1")
if use_dynamic_attention:
from .attention import dynamic_scaled_dot_product_attention