Fastai v0.7 install issues thread

Chandrak · February 13, 2019, 7:03pm

!pip install fastai==0.7.0
!pip install torchtext==0.2.3
from fastai.text import *
import html

I get Dataset not defined error.

NameError Traceback (most recent call last)
in
----> 1 from fastai.text import *
2 import html

~/notebooks/notebooks/fastai/courses/dl2/fastai/text.py in
1 from .core import *
----> 2 from .learner import *
3 from .lm_rnn import *
4 from torch.utils.data.sampler import Sampler
5 import spacy

~/notebooks/notebooks/fastai/courses/dl2/fastai/learner.py in
4 from .transforms import *
5 from .model import *
----> 6 from .dataset import *
7 from .sgdr import *
8 from .layer_optimizer import *

~/notebooks/notebooks/fastai/courses/dl2/fastai/dataset.py in
185 return full_names, label_arr, all_labels
186
–> 187 class BaseDataset(Dataset):
188 “”“An abstract class representing a fastai dataset. Extends torch.utils.data.Dataset.”""
189 def init(self, transform=None):

NameError: name ‘Dataset’ is not defined

How to resolve this? Can someone please help me?
Thank you,

ashis · March 10, 2019, 5:21am

Hi friends,
Recently, I came across many problems after installing fastai v 0.7.0 for Part 2 of the course.
Unfamiliar Error when Running learn.fit().
Did any of you folks came across such problem and managed to get a solution for this.
Really appreciate your efforts.
Thanks,
Ashis

francisc · March 30, 2019, 8:12pm

Hi, I’m having the same error you talked about in here, and the symlink seems fine as well. Did you ever figure out what was the problem?

mothas · April 17, 2019, 7:01pm

I forget the specifics now. But I think I was not running the conda environment. It was a simple fix like that.

thinkortrap · April 18, 2019, 3:07pm

Hello, Im having issues downloading fastai. I’m getting this error but i did download pytoch

chloe2018 · May 3, 2019, 11:30am

fastai 0.7.0 not working Google Colaboratory anymore?

I am using Google Colaboratory and fastai 0.7.0

It worked fine some time back (I would guess in January this year),
now I am having a look at the notebooks, and they fail to run (without any code changes)

I have been using the dogs vs cats example from the course.

!pip install -q torchtext==0.2.3
!pip install -q fastai==0.7.0

torch.version
gives 0.3.1

arch=resnet34
data = ImageClassifierData.from_paths(PATH, tfms=tfms_from_model(arch, sz))
learn = ConvLearner.pretrained(arch, data, precompute=True)
learn.fit(0.01, 2)

module ‘torch.nn.init’ has no attribute ‘kaiming_normal_’

No idea why it worked some months back and what changed.

I try to figure out what the issue is, and it seems to be due a need to upgrade to pytorch:

I kill the Colab seesions, and reset try to upgrade
!pip install -q http://download.pytorch.org/whl/{accelerator}/torch-0.4.1-{platform}-linux_x86_64.whl torchvision

torch.version
gives 0.4.1

Running the code again
arch=resnet34
data = ImageClassifierData.from_paths(PATH, tfms=tfms_from_model(arch, sz))
learn = ConvLearner.pretrained(arch, data, precompute=True)
learn.fit(0.01, 2)

/usr/local/lib/python3.6/dist-packages/torch/cuda/init.py:114: UserWarning:
Found GPU0 Tesla T4 which requires CUDA_VERSION >= 9000 for
optimal performance and fast startup time, but your PyTorch was compiled
with CUDA_VERSION 8000. Please install the correct PyTorch binary
using instructions from http://pytorch.org

warnings.warn(incorrect_binary_warn % (d, name, 9000, CUDA_VERSION))

so I think it is runs on CPU now instead, and it never finishes (it runs out of memory on colab after a while)

If I try to upgrade to pytorch 1.1 I get

ERROR: fastai 0.7.0 has requirement torch<0.4, but you’ll have torch 1.1.0 which is incompatible.

Anyone successfully running 0.7.0 in Colab now?

gilfoyle · May 4, 2019, 8:12am

After creating the environment…when trying to activate by "
conda activate fastai-cpu"
I am getting below error
CommandNotFoundError: Your shell has not been properly configured to use ‘conda activate’.
To initialize your shell, run

$ conda init <SHELL_NAME>

mdeore · May 5, 2019, 2:01am

I have created docker container for FastAI version 0.7.0. You can runn following command to pull the container:

docker pull mdeore/fastai_0_7_0

Hope this helps.

RuAlice · July 3, 2019, 12:35pm

I’m also having problems with torch version, did you manage to sort this?

I’m getting this:
torchvision 0.3.0 has requirement torch>=1.1.0, but you’ll have torch 0.1.2.post2 which is incompatible.

And This:
Command “/Applications/anaconda3/bin/python -u -c “import setuptools, tokenize;file=’/private/var/folders/b2/l0wpqyh966759ms4lr29d9cm0000gn/T/pip-install-w360hnca/bcolz/setup.py’;f=getattr(tokenize, ‘open’, open)(file);code=f.read().replace(’\r\n’, ‘\n’);f.close();exec(compile(code, file, ‘exec’))” install --record /private/var/folders/b2/l0wpqyh966759ms4lr29d9cm0000gn/T/pip-record-envqcys5/install-record.txt --single-version-externally-managed --compile” failed with error code 1 in /private/var/folders/b2/l0wpqyh966759ms4lr29d9cm0000gn/T/pip-install-w360hnca/bcolz/

luskmusk · July 20, 2019, 4:51pm

Ruth,

Any chance you were able to resolve this? I’m running into same issue.

draz · August 23, 2019, 12:40am

@mdeore Does it run on a gpu or cpu env? Thanks!

mdeore · August 23, 2019, 3:10am

@draz its for CPU but will make it GPU as well.

draz · August 23, 2019, 5:09am

Hello FastAI friends,

I think that I am at the very right topic to get helped on my problem. I am trying to create a docker on Ubuntu for fastai-cpu v0.7. I have created the following Dockerfile:

FROM ubuntu:18.04
ENV PYTHON_VERSION=3.6

WORKDIR /fastaidir

ENV TINI_VERSION v0.18.0

ADD https://github.com/krallin/tini/releases/download/${TINI_VERSION}/tini /tini
ADD https://github.com/etheld/kubevaultenv/releases/download/0.1.12/kubevaultenv-linux-amd64 /vaultenv

RUN apt-get update && \
    apt-get install -y git wget curl unzip libsm6 libxrender1 && \
    curl -o ~/miniconda.sh -O  https://repo.continuum.io/miniconda/Miniconda3-4.5.11-Linux-x86_64.sh  && \
    chmod +x ~/miniconda.sh && \
    ~/miniconda.sh -b -p /opt/conda && \
    rm ~/miniconda.sh && \
    /opt/conda/bin/conda install conda-build && \
    git clone https://github.com/fastai/fastai.git && \
    cd /fastaidir/fastai && ls && /opt/conda/bin/conda env create -f environment-cpu.yml -n fastai-cpu && \
    /opt/conda/bin/conda clean -ya && \
    chmod +x /tini && \
    chmod +x /vaultenv && \
    ln -s /data /fastaidir/data

ADD config_fastai_cpu.sh /
ENV USER fastai

ENTRYPOINT ["/tini", "--"]
CMD ["/config_fastai_cpu.sh”]

, where at the end I am configuring the fastai-cpu with the config_fastai_cpu.sh, inspired from Jeremy at the beginning of this thread:

#!/bin/bash

source /opt/conda/etc/profile.d/conda.sh
conda activate fastai-cpu

pip uninstall -y matplotlib  ;
python -m pip install --upgrade pip  ;
pip install matplotlib  ;
pip install fastai==0.7.0  ;
pip install torchtext==0.2.3  ;
pip install flask==1.0.2    ;

exec /opt/conda/envs/fastai-cpu/bin/python -c 'from fastai.conv_learner import *; print("\nimported \n")'
printf "%*s\n" $ColumnsN |tr " " “_”

This was a working code about a year ago. I bet it’s something about the versions of torch, (or Ubuntu?, or Miniconda?), and at the very end I am getting this error:

Traceback (most recent call last):
  File "<string>", line 1, in <module>
  File "/opt/conda/envs/fastai-cpu/lib/python3.6/site-packages/fastai/conv_learner.py", line 1, in <module>
    from .core import *
  File "/opt/conda/envs/fastai-cpu/lib/python3.6/site-packages/fastai/core.py", line 2, in <module>
    from .torch_imports import *
  File "/opt/conda/envs/fastai-cpu/lib/python3.6/site-packages/fastai/torch_imports.py", line 2, in <module>
    import torch, torchvision, torchtext
  File "/opt/conda/envs/fastai-cpu/lib/python3.6/site-packages/torch/__init__.py", line 56, in <module>
    from torch._C import *
ImportError: libcudart.so.8.0: cannot open shared object file: No such file or directory

I don’t understand what I am missing. Please suggest. Thanks!

banksiaboy · August 23, 2019, 6:39am

OMG it works!
I spent hours trying to follow the various install documents. Please do this document in just one place, and replace all the other awful ones. No greater barrier to a software community than an impossible setup.
Thanks…

minh · August 24, 2019, 9:10am

I wrote a document on How to Install Fastai v0.7 on AWS for Fastai Machine Learning Course that details every installation step to get you started on the Fastai ML course. I agree with @banksiaboy that there is no greater barrier to software development and learning than an impossible setup.

draz · August 25, 2019, 12:31am

@minh I think that the step 3 is not correct. By doing git clone https://github.com/fastai/fastai.git it basically installs v1.x, not v0.7.

minh · August 25, 2019, 1:20pm

I tried and got fasai v0.7 installed correctly. Did you somehow get v1 installed?

draz · August 25, 2019, 4:47pm

On a brand new system (created with Docker) v1 is installed in this way. Moreover, the instructions in https://github.com/fastai/fastai/tree/master/old describe exactly what I am saying.

DebabrataRoy · August 25, 2019, 5:13pm

I have installed latest fastai version. But i am not able to run the ML course. How can i access 0.7version now. Does all the functions of 0.7 version is present in current version?

draz · August 25, 2019, 9:28pm

@DebabrataRoy The latest version is different from the v0.7. You should follow the instructions in https://github.com/fastai/fastai/tree/master/old that refer to the old version (v0.7).