mirror of
https://github.com/NVIDIA/TensorRT-LLM.git
synced 2026-02-07 19:51:50 +08:00
34 lines
759 B
Python
34 lines
759 B
Python
import pytest
|
|
import torch
|
|
|
|
from tensorrt_llm._torch.hostfunc import HOSTFUNC_USER_DATA_HANDLES, hostfunc
|
|
|
|
|
|
@pytest.mark.skip(reason="https://nvbugs/5643631")
|
|
def test_hostfunc():
|
|
|
|
@hostfunc
|
|
def increase(x: torch.Tensor):
|
|
x.add_(1)
|
|
|
|
x = torch.zeros(10, dtype=torch.int32)
|
|
|
|
stream = torch.cuda.Stream()
|
|
with torch.cuda.stream(stream):
|
|
for _ in range(5):
|
|
increase(x)
|
|
|
|
g = torch.cuda.CUDAGraph()
|
|
with torch.cuda.graph(g, stream=stream):
|
|
increase(x)
|
|
increase(x)
|
|
torch.cuda.synchronize()
|
|
|
|
with torch.cuda.stream(stream):
|
|
for _ in range(10):
|
|
g.replay()
|
|
|
|
torch.cuda.synchronize()
|
|
assert (x == 25).all().item()
|
|
assert len(HOSTFUNC_USER_DATA_HANDLES) == 2
|