"NameError: name 'T' is not defined" Deep Learning Part 2(dl2) Lesson 10 IMDB

Hi,

I am trying to run imdb.ipynb(from dl2 lesson 10) and I have reached the ‘wikitext103 conversion’ topic’s last cell.

wgts['0.encoder.weight'] = T(new_w)
wgts['0.encoder_with_dropout.embed.weight'] = T(np.copy(new_w))
wgts['1.decoder.weight'] = T(np.copy(new_w))

But it gives this error:

NameError                                 Traceback (most recent call last)
<ipython-input-294-678a783a7cd5> in <module>()
----> 1 wgts['0.encoder.weight'] = T(new_w)
      2 wgts['0.encoder_with_dropout.embed.weight'] = T(np.copy(new_w))
      3 wgts['1.decoder.weight'] = T(np.copy(new_w))

NameError: name 'T' is not defined

Basically, it can’t find the function T()

Here is all what I have imported:
from fastai.text import *
import html
import os
from pathlib import Path
import pandas as pd
import numpy as np
import scipy
import sklearn
import sklearn.model_selection
import re
import torch

What all have I tried doing?
I have searched about this on the internet but I can’t find anything about the function “T()”

I came to know about numpy.ndarray.T which is an Attribute which transposes the ndarray but there is no function “T()” anywhere.

I also searched in fastai repo using git search but I couldn’t find it. I think it’s a function from fastai or torch and I am not sure what exactly it does.

Can somebody please help?

Hey @jeremy, Sorry to tag you on this but I really tried searching the answer today too and couldn’t find it. If you could please take some time to look at this once?

Sorry, its really urgent for me.

I think you’re looking for “import torch.tensor as T” ?

1 Like

Hey @claeyzre, Thanks so much! I am past that now but again stuck at this one:
trn_dl = LanguageModelLoader(np.concatenate(trn_lm), bs, bptt)
val_dl = LanguageModelLoader(np.concatenate(val_lm), bs, bptt)
md = LanguageModelData(PATH, 1, vs, trn_dl, val_dl, bs=bs, bptt=bptt)

AttributeError                            Traceback (most recent call last)
<ipython-input-111-12702c19051f> in <module>()
----> 1 trn_dl = LanguageModelLoader(np.concatenate(trn_lm), bs, bptt)
      2 val_dl = LanguageModelLoader(np.concatenate(val_lm), bs, bptt)
      3 md = LanguageModelData(PATH, 1, vs, trn_dl, val_dl, bs=bs, bptt=bptt)

/usr/local/lib/python3.6/dist-packages/fastai/text/data.py in __init__(self, dataset, bs, bptt, backwards, shuffle, max_len)
    211         self.dataset,self.bs,self.bptt,self.backwards,self.shuffle = dataset,bs,bptt,backwards,shuffle
    212         self.first,self.i,self.iter = True,0,0
--> 213         self.n = len(np.concatenate(dataset.x)) // self.bs
    214         self.max_len,self.num_workers = max_len,0
    215 

AttributeError: 'numpy.ndarray' object has no attribute 'x'

I don’t get what ‘x’ is here, any ideas?

Thanks again for telling about the T function :slight_smile:

You are using a wrong version of fastai,
Use this installation guide if you want to make it work correctly Fastai v0.7 install issues thread

Hi @aashish1111, I stuck in the same place as you. Can you help me how did you solved it??

I’m using 1.0.40.dev0 and I’m getting the same error as well. I’ve tried installing using conda, pip and eventually installed by source but same error.

Did you resolve this?

No @sabzo I was not able to

Hi @aashish1111, I stuck in the same place as you. Can you help me how did you solved it??

1 Like

same! were you able to solve it ?

@jayrodge @sabzo I had the same error! Got my notebook running and documented all the steps here https://link.medium.com/MtMCa964TU . Hope this helps!

1 Like