Missing requirements

[goodmami]

After following the instructions in the README, I find there are some missing Python dependencies:

• torchtext
• configargparse

The latter package is part of the error in the second screenshot of #9, as well. Maybe you could provide a pip freeze > requirements.txt file? Here's what my virtual environment has currently:

$ pip freeze
boto3==1.12.34
botocore==1.15.34
certifi==2019.11.28
chardet==3.0.4
click==7.1.1
ConfigArgParse==1.1
docutils==0.15.2
filelock==3.0.12
idna==2.9
jmespath==0.9.5
joblib==0.14.1
numpy==1.18.2
Pillow==7.1.0
pkg-resources==0.0.0
python-dateutil==2.8.1
regex==2020.2.20
requests==2.23.0
s3transfer==0.3.3
sacremoses==0.0.38
sentencepiece==0.1.85
six==1.14.0
tokenizers==0.5.2
torch==1.4.0+cpu
torchtext==0.5.0
torchvision==0.5.0+cpu
tqdm==4.44.1
transformers==2.7.0
urllib3==1.25.8

Not all of these are relevant to all users, though, particularly the +cpu ones.

[loic-vial]

Hello, thank you for your post !

I don't have much time currently to work on this repository, but yes, I will certainly update the Readme and write a requirements.txt file, in two/three months from now.

For installing PyTorch, I prefer to advice people to install it from Conda, or via a system-wide package manager (apt-get, pacman...), because it will also install the CUDA library if you are working on GPU, or automatically install the CPU version is there is no GPU available.

For the other packages, you're right about "torchtext" and "configargparse", on some computers I had to manually install them after the other dependencies, but I have to check and clean first the other dependencies. For instance :

• definitely change the "pytorch_pretrained_bert" dependency in "transformers" (pytorch_pretrained_bert vs transformers #16 )
• remove/update the dependenciy to the "onmt" package of the project: it's a copy/paste of some classes from the ONMT tool, from the time when they implemented the TransformerEncoder and it was not available "natively" in PyTorch.

I keep this issue open to remind me to do all that (after I finish my thesis 😁 )