tissue333
  • Introduction
  • JPEG and Neural Network
    • Progress
    • TodoList
  • Findings
    • Why inconsistensy in backward propagation (pytorch)?
    • GOMP_4.0 not found
    • How to support parallel computing with constant tensor in a self-defined layer?
  • Seashell
    • Review
    • Spatial
    • Dahlia Programmability
    • FuTil
    • FuTil Paper Review
  • Configurable DNN to FPGA
    • General Idea
    • Event Queue Dialect (MLIR)
  • Progress Log
  • Random
    • ubuntu
Powered by GitBook
On this page

Was this helpful?

  1. Findings

GOMP_4.0 not found

PreviousWhy inconsistensy in backward propagation (pytorch)?NextHow to support parallel computing with constant tensor in a self-defined layer?

Last updated 6 years ago

Was this helpful?

One day I was trying to downsample an image with torch.nn.functional.interpolate, pytorch reports me that

[AttributeError: module ‘torch.nn.functional’ has no attribute 'interpolate']

This is strange because the repo says has already been implemented.

In an discussion I found this function is introduced in 0.4.1, and my pytorch version is 0.4.0.

To get update to latest version, conda update pytorch torchvision is not enough though. It will only keep your pytorch version at 4.0.

One should adds config before updating the pytorch.

conda config --add channels soumith

Updating doesn't solve the problem. Not all libraries get updated.

ImportError: /path/to/python/site-packages/torch/lib/libgomp.so.1: version `GOMP_4.0' not found (required by /path/to/python/site-packages/torch/lib/libTH.so.1)

Let's fix this.

rm path/to/python/site-packages/torch/lib/libgomp.so.1
ln -s /usr/lib/libgomp.so.1 path/to/python/site-packages/torch/lib/libgomp.so.1
interpolate
issue