Transformers library version

[rezash1]

Hi,
Can you please add a requirements.txt file for easier reproducibility?
Thanks

[AndriyMulyar]

That is not necessary. It is a package which you can pip install. You can find all requirements in the setup.py file.

…

On Mon, Jun 8, 2020, 5:27 PM Reza (Amirreza) Shirani < ***@***.***> wrote: Hi, Can you please add a requirements.txt file for easier reproducibility? Thanks — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#1>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ADJ4TBUOCW6BYFYTZSHGKKDRVVJVBANCNFSM4NYZDFPQ> .

[rezash1]

Thanks!
But it doesn't contain transformers and torch versions.

[AndriyMulyar]

Thanks for the heads up! It should work with the current versions of both packages. Could you verify and PR a setup.py with the current version restricted. Thanks!

…

On Mon, Jun 8, 2020, 5:45 PM Reza (Amirreza) Shirani < ***@***.***> wrote: Thanks! But it doesn't contain transformers and torch versions. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#1 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ADJ4TBQE6JP7ZD5WTHRX6V3RVVLWRANCNFSM4NYZDFPQ> .

[rezash1]

I get the following error when I run predict_ner.py:
TypeError: init() got an unexpected keyword argument 'add_special_tokens'

These are the versions I'm using:
torch==1.5.0
transformers==2.11.0

[AndriyMulyar]

That file utilizes the 'tokenizers' package not 'transformers' package for tokenization.

…

On Mon, Jun 8, 2020, 5:59 PM Reza (Amirreza) Shirani < ***@***.***> wrote: I get the following error when I run predict_ner.py: TypeError: *init*() got an unexpected keyword argument 'add_special_tokens' These are the versions I'm using: torch==1.5.0 transformers==2.11.0 — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#1 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ADJ4TBSDEXSV6HS6AMKDEFLRVVNM3ANCNFSM4NYZDFPQ> .

[AndriyMulyar]

Updated setup.py to reflect - where is ci/cd when you need it. Please let me know if there are any other ways that friction to reproducibility can be reduced.

[rezash1]

Thanks! I think there is still a mismatch between my version of 'tokenizers' and this implementation. so I removed 'add_special_tokens=False' and 'encoding.original_str' in 'predict_ner.py' which gave me an error.

[AndriyMulyar]

I updated the setup.py and got a clean build from it working a few days back. Try setting up a virtual environment and 'pip install' ing the latest version from master. Thanks!

…

On Fri, Jun 19, 2020, 1:55 PM Reza (Amirreza) Shirani < ***@***.***> wrote: Thanks! I think there is still a mismatch between my version of 'tokenizers' and this implementation. so I removed 'add_special_tokens=False' and 'encoding.original_str' in 'predict_ner.py' which gave me an error. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#1 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ADJ4TBRT3W3AFHVPESB2URTRXOQ75ANCNFSM4NYZDFPQ> .