AttributeError: module 'torchtext' has no attribute 'transforms'

[nqbao]

Describe the bug
I got this error while trying to train a model with encoder = bert

To Reproduce

trainer:
    validation_field: engagement_label
    validation_metric: roc_auc
    early_stop: 10

input_features:
    - name: text1
      type: text
      encoder: 
          type: bert
    - name: text2
      type: text
      encoder: 
          type: bert

output_features:
    - name: engagement_label
      type: binary
      threshold: 0.6

Expected behavior
model is trained with no error

Screenshots
n/a

Environment (please complete the following information):

• OS: [e.g. iOS]
• Version [e.g. 22]
• Python 3.11.8
• ludwig==0.10.3

Additional context
Add any other context about the problem here.

[nqbao]

this seems to be relate to #4013

[mhabedank]

Hi @nqbao can you give us some deeper insights. The trace and some test data would be helpful.

[mhabedank]

@skanjila @ethanreidel we need to discuss how to deal with torchtext in general. The library will not be further developed. https://pytorch.org/text/stable/

[nqbao]

@mhabedank i don't remember how to reproduce this one but basically this is an incompatibility with pytorch and pytorch text, so i don't know which version to install to make this works

it would be nice if we specify somewhere what are the requirements

[mhabedank]

@nqbao we worked on this, can you install torchtext 0.17.2 and try again?