for in range(50):
= model(random_input1, random_input2)
测速
times = torch.zeros(iterations) # 存储每轮iteration的时间
with torch.nograd():
for iter in range(iterations):
starter.record()
= model(random_input1, random_input2)
ender.record()
mean_time = times.mean().item()
print("Inference time: {:.6f}, FPS: {} ".format(mean_time, 1000/mean_time))
And the error is:
Traceback (most recent call last):
File "inferencetime.py", line 17, in = model(random_input1, random_input2)
File "/root/anaconda3/envs/ztenv/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1501, in _call_impl
return forward_call(*args, *kwargs)
File "/mnt/zt/VcT/VcT_code/models/networks.py", line 364, in forward
A = self.knngraph(x3) # b hw hw
File "/mnt/zt/VcT/VcT_code/models/networks.py", line 352, in knngraph
d = torch.sparse.FloatTensor(indice1, vals, torch.Size([b n, n])).to_dense()
RuntimeError: CUDA error: device-side assert triggered
CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.
For debugging consider passing CUDA_LAUNCH_BLOCKING=1.
Compile with TORCH_USE_CUDA_DSA to enable device-side assertions.
`
hello,we have met some issues when computing inference time. Below is th code we used. ` import torch from models.networks import Reliable_Transformer
iterations = 300 # 重复计算的轮次
model = Reliable_Transformer(input_nc=3, output_nc=2, resnet_stages_num=4, with_pos='learned', enc_depth=1, dec_depth=8) device = torch.device("cuda:0") model.to(device)
random_input1 = torch.randn(1, 3, 256, 256).to(device) random_input2 = torch.randn(1, 3, 256, 256).to(device) starter, ender = torch.cuda.Event(enable_timing=True), torch.cuda.Event(enable_timing=True)
GPU预热
for in range(50): = model(random_input1, random_input2)
测速
times = torch.zeros(iterations) # 存储每轮iteration的时间 with torch.nograd(): for iter in range(iterations): starter.record() = model(random_input1, random_input2) ender.record()
同步GPU时间
mean_time = times.mean().item() print("Inference time: {:.6f}, FPS: {} ".format(mean_time, 1000/mean_time))
= model(random_input1, random_input2)
File "/root/anaconda3/envs/ztenv/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1501, in _call_impl
return forward_call(*args, *kwargs)
File "/mnt/zt/VcT/VcT_code/models/networks.py", line 364, in forward
A = self.knngraph(x3) # b hw hw
File "/mnt/zt/VcT/VcT_code/models/networks.py", line 352, in knngraph
d = torch.sparse.FloatTensor(indice1, vals, torch.Size([b n, n])).to_dense()
RuntimeError: CUDA error: device-side assert triggered
CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.
For debugging consider passing CUDA_LAUNCH_BLOCKING=1.
Compile with
And the error is:Traceback (most recent call last): File "inferencetime.py", line 17, inTORCH_USE_CUDA_DSAto enable device-side assertions. `Can you help us?