ValueError: len() should return >= 0 with LPPool2d operator

[Thrsu]

Description:

The error message is as below:

Traceback (most recent call last):
  ...
    model_trt = torch2trt(model, [input_data])
  File "/root/miniconda3/envs/nnsmith/lib/python3.9/site-packages/torch2trt-0.4.0-py3.9.egg/torch2trt/torch2trt.py", line 778, in torch2trt
    outputs = module(*inputs)
  File "/root/miniconda3/envs/nnsmith/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1518, in _wrapped_call_impl
    return self._call_impl(*args, **kwargs)
  File "/root/miniconda3/envs/nnsmith/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1568, in _call_impl
    result = forward_call(*args, **kwargs)
  File "/root/miniconda3/envs/nnsmith/lib/python3.9/site-packages/torch/nn/modules/pooling.py", line 984, in forward
    return F.lp_pool2d(input, float(self.norm_type), self.kernel_size,
  File "/root/miniconda3/envs/nnsmith/lib/python3.9/site-packages/torch2trt-0.4.0-py3.9.egg/torch2trt/torch2trt.py", line 300, in wrapper
    outputs = method(*args, **kwargs)
  File "/root/miniconda3/envs/nnsmith/lib/python3.9/site-packages/torch/nn/functional.py", line 1042, in lp_pool2d
    return (torch.sign(out) * relu(torch.abs(out))).mul(kw * kh).pow(1.0 / norm_type)
  File "/root/miniconda3/envs/nnsmith/lib/python3.9/site-packages/torch2trt-0.4.0-py3.9.egg/torch2trt/torch2trt.py", line 309, in wrapper
    converter["converter"](ctx)
  File "/root/miniconda3/envs/nnsmith/lib/python3.9/site-packages/torch2trt-0.4.0-py3.9.egg/torch2trt/converters/mul.py", line 15, in convert_mul
    input_a_trt, input_b_trt = broadcast_trt_tensors(ctx.network, [input_a_trt, input_b_trt], len(output.shape))
  File "/root/miniconda3/envs/nnsmith/lib/python3.9/site-packages/torch2trt-0.4.0-py3.9.egg/torch2trt/torch2trt.py", line 193, in broadcast_trt_tensors
    if len(t.shape) < broadcast_ndim:
ValueError: __len__() should return >= 0

Reproduce:

Here is a minimal script to reproduce the issue:

import torch
from torch.nn import Module
from torch2trt import torch2trt

model = torch.nn.LPPool2d(2, 2, 2,).eval().cuda()
input_data = torch.randn([1, 3, 7, 7], dtype=torch.float32).cuda()
model_trt = torch2trt(model, [input_data])

Environment

• torch: 2.1.1
• torch2trt: 0.4.0
• tensorrt: 8.6.1

[farzanehnakhaee70]

Hi,
Is there any update for this issue?

[hongliyu0716]

The ReLU6 operator has the same problem, here is the script:

import torch
from torch.nn import Module
from torch2trt import torch2trt

model = torch.nn.ReLU6().eval().cuda()
input_data=torch.randint(1, 100, [4, 10], dtype=torch.uint8).cuda()
model_trt = torch2trt(model, [input_data])

[Qingchao Shen]

I came across the same issue when using InstanceNorm2d.