NameError: global name 'self' is not defined

Subhobrata_Mukharjee · February 28, 2019, 4:16am

Hi All,

Can someone please help me with the below error.

Python is 2.7 and I am using pytorch in colab.

I am getting below error.

NameError: global name ‘self’ is not defined

The code is bit long so I have given Github link.

github.com

subhobrata/deep-learning-for-visual-computing/blob/master/Lecture_19.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Lecture 19.ipynb",
      "version": "0.3.2",
      "provenance": [],
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python2",
      "display_name": "Python 2"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {

This file has been truncated. show original

Thanks and Regards,
Subho

ptrblck · February 28, 2019, 12:20pm

L1Penalty uses self.l1weight which is unknown. Probably you wanted to use ctx.l1weight instead.

Subhobrata_Mukharjee · February 28, 2019, 3:09pm

Hi ptrblck,

I tried, but now I am getting below error.

RuntimeError: function L1PenaltyBackward returned an incorrect number of gradients (expected 2, got 1)

Colab link

github.com

subhobrata/deep-learning-for-visual-computing/blob/master/Lecture_19_2.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Lecture 19_2.ipynb",
      "version": "0.3.2",
      "provenance": [],
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python2",
      "display_name": "Python 2"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {

This file has been truncated. show original

Thanks and Regards,
Subho

Subhobrata_Mukharjee · March 1, 2019, 3:56am

Hi All,

Please help me on this.

Thanks and Regards,
Subho

ptrblck · March 2, 2019, 4:16pm

Hi Subho,

have a look at the Extending PyTorch docs.
From these docs:

It should return as many Tensor s as there were inputs, with each of them containing the gradient w.r.t. its corresponding input. If your inputs didn’t require gradient ( needs_input_grad is a tuple of booleans indicating whether each input needs gradient computation), or were non- Tensor objects, you can return None .

As it seems you are working on some kind of lecture assignment, you shouldn’t expect to get a complete solution.
Also, I’m not a big fan of posting repeatedly and would l like to ask you to be a bit more patient

Subhobrata_Mukharjee · March 2, 2019, 4:24pm

ok sure Thanks ptrblck,