Cannot use ConveRT featurizer


I wanted to use the new pipeline “pretrained_embeddings_convert” which includes ConveRT, but I’m getting this error:

Exception: Not all required importable packages are installed. To use this pipeline, you need to install the missing dependencies. Please install the package(s) that contain the module(s): tensorflow_text

I then tried to install tensorflow_text which I believe is included in tensorflow-text package (as it says here), but I got the error:

ERROR: Could not find a version that satisfies the requirement tensorflow-text (from versions: none)
ERROR: No matching distribution found for tensorflow-text

Any idea?

Thank you, Tiziano

Hi @tiziano,

are you operating on Windows? Currently, tensorflow_text is not supported on non-Linux platforms and as far as I have seen, there is no plan to support it during the next few weeks.

If you need/want to try, try the WSL for Ubuntu or simply use a dockerized Rasa.

If you are using Linux, could you please post the commands you used to install?

Regards Julian

1 Like

Hi @JulianGerhard,

Thank you for your answer. Yes I’m using Windows. I will try using WSL for Ubuntu or dockerized Rasa, thank you

That mean we cannot use ConveRTTokenizer and ConveRTFeaturizer in windows ?

Hi @tiziano

I have faced the same issue. It seems to be associated to the tokenization. To solve it I have replaced the ConveRTTokenizer to WhitespaceTokenizer.

Replace the code in your config.yml file to:

language: en
  - name: WhitespaceTokenizer
  - name: RegexFeaturizer
  - name: LexicalSyntacticFeaturizer
  - name: CountVectorsFeaturizer
  - name: CountVectorsFeaturizer
analyzer: "char_wb"
min_ngram: 1
max_ngram: 4
  - name: DIETClassifier
epochs: 100
  - name: EntitySynonymMapper
  - name: ResponseSelector
epochs: 100
1 Like