biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[Hardware][Intel GPU] Add Intel GPU(XPU) inference backend (#3814)

Browse Source 
Co-authored-by: Jiang Li <jiang1.li@intel.com>
Co-authored-by: Abhilash Majumder <abhilash.majumder@intel.com>
Co-authored-by: Abhilash Majumder <30946547+abhilash1910@users.noreply.github.com>
This commit is contained in:
Kunshang Ji
2024-06-18 02:01:25 +08:00
committed by GitHub
parent 1f12122b17
commit 728c4c8a06
31 changed files with 1998 additions and 24 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
vllm/model_executor/custom_op.py
View File

@@ -1,6 +1,6 @@
import torch.nn as nn
from vllm.utils import is_cpu, is_hip, is_tpu
from vllm.utils import is_cpu, is_hip, is_tpu, is_xpu
class CustomOp(nn.Module):
@@ -29,9 +29,7 @@ class CustomOp(nn.Module):
return self.forward_cuda(*args, **kwargs)
def forward_xpu(self, *args, **kwargs):
# By default, we assume that XPU ops are compatible with CUDA ops.
# NOTE(woosuk): This is a placeholder for future extensions.
return self.forward_cuda(*args, **kwargs)
raise NotImplementedError
def forward_cpu(self, *args, **kwargs):
# By default, we assume that CPU ops are compatible with CUDA ops.
@@ -58,5 +56,7 @@ class CustomOp(nn.Module):
return self.forward_cpu
elif is_tpu():
return self.forward_tpu
elif is_xpu():
return self.forward_xpu
else:
return self.forward_cuda