I am trying the pytorch-quantization (TensorRT/tools/pytorch-quantization at release/10.3 · NVIDIA/TensorRT · GitHub), and encountered a problem, hope some one can give me some advice on fixing it.
pytorch_quantization version: release/10.3
Installation from source code:
git clone GitHub - NVIDIA/TensorRT: NVIDIA® TensorRT™ is an SDK for high-performance deep learning inference on NVIDIA GPUs. This repository contains the open source components of TensorRT.
cd TensorRT/tools/pytorch-quantization
python setup.py install
Running:
python -c “import pytorch_quantization”
Report following error:
Traceback (most recent call last):
File “”, line 1, in
File “/data/theoWS/nvidia-quant/TensorRT/tools/pytorch-quantization/pytorch_quantization/init.py”, line 20, in
from .quant_modules import *
File “/data/theoWS/nvidia-quant/TensorRT/tools/pytorch-quantization/pytorch_quantization/quant_modules.py”, line 23, in
from pytorch_quantization import nn as quant_nn
File “/data/theoWS/nvidia-quant/TensorRT/tools/pytorch-quantization/pytorch_quantization/nn/init.py”, line 19, in
from pytorch_quantization.nn.modules.tensor_quantizer import *
File “/data/theoWS/nvidia-quant/TensorRT/tools/pytorch-quantization/pytorch_quantization/nn/modules/tensor_quantizer.py”, line 24, in
from pytorch_quantization.tensor_quant import QuantDescriptor, tensor_quant, fake_tensor_quant, scaled_e4m3
File “/data/theoWS/nvidia-quant/TensorRT/tools/pytorch-quantization/pytorch_quantization/tensor_quant.py”, line 28, in
from pytorch_quantization import cuda_ext
ImportError: cannot import name ‘cuda_ext’ from partially initialized module ‘pytorch_quantization’ (most likely due to a circular import) (/data/theoWS/nvidia-quant/TensorRT/tools/pytorch-quantization/pytorch_quantization/init.py)
What’s the problem and how to fix it? Thank you.