Rasa 3.0 Error on train model (LanguageModelFeaturizer , bert)

Rasa Open Source
1

I got error while I train model by using LanguageModelFeaturizer with bert. Which success on Rasa 2.x. Anyone know the solution? Thanks.

Here’s the information

All the layers of TFBertModel were initialized from the model checkpoint at rasa/LaBSE.
If your task is similar to the task the model of the checkpoint was trained on, you can already use TFBertModel for predictions without further training.
Traceback (most recent call last):
  File "/home/hko/jonathan/venv2/lib/python3.7/site-packages/rasa/engine/graph.py", line 461, in __call__
    output = self._fn(self._component, **run_kwargs)
  File "/home/hko/jonathan/venv2/lib/python3.7/site-packages/rasa/nlu/featurizers/dense_featurizer/lm_featurizer.py", line 731, in process_training_data
    batch_docs = self._get_docs_for_batch(batch_messages, attribute)
  File "/home/hko/jonathan/venv2/lib/python3.7/site-packages/rasa/nlu/featurizers/dense_featurizer/lm_featurizer.py", line 689, in _get_docs_for_batch
    batch_token_ids, batch_tokens, batch_examples, attribute, inference_mode
  File "/home/hko/jonathan/venv2/lib/python3.7/site-packages/rasa/nlu/featurizers/dense_featurizer/lm_featurizer.py", line 620, in _get_model_features_for_batch
    batch_attention_mask, padded_token_ids
  File "/home/hko/jonathan/venv2/lib/python3.7/site-packages/rasa/nlu/featurizers/dense_featurizer/lm_featurizer.py", line 470, in _compute_batch_sequence_features
    np.array(padded_token_ids), attention_mask=np.array(batch_attention_mask)
  File "/home/hko/jonathan/venv2/lib/python3.7/site-packages/keras/engine/base_layer.py", line 1037, in __call__
    outputs = call_fn(inputs, *args, **kwargs)
  File "/home/hko/jonathan/venv2/lib/python3.7/site-packages/transformers/models/bert/modeling_tf_bert.py", line 1143, in call
    training=inputs["training"],
  File "/home/hko/jonathan/venv2/lib/python3.7/site-packages/keras/engine/base_layer.py", line 1037, in __call__
    outputs = call_fn(inputs, *args, **kwargs)
  File "/home/hko/jonathan/venv2/lib/python3.7/site-packages/transformers/models/bert/modeling_tf_bert.py", line 803, in call
    attention_mask_shape = shape_list(inputs["attention_mask"])
  File "/home/hko/jonathan/venv2/lib/python3.7/site-packages/transformers/modeling_tf_utils.py", line 1831, in shape_list
    static = tensor.shape.as_list()
AttributeError: 'tuple' object has no attribute 'as_list'

pipeline:
   - name: "SpacyNLP"
     model: "zh_core_web_trf"
   - name: "SpacyTokenizer"
   - name: RegexFeaturizer
   - name: LexicalSyntacticFeaturizer
   - name: CountVectorsFeaturizer
   - name: CountVectorsFeaturizer
     analyzer: char_wb
     min_ngram: 1
     max_ngram: 4
   - name: LanguageModelFeaturizer
     model_name: "bert"
     #model_name: "bert"
     #model_weights: "rasa/LaBSE"
     #model_weights: "bert-base-multilingual-uncased"
     cache_dir: null     
   - name: DIETClassifier
     epochs: 100
   - name: EntitySynonymMapper
   - name: ResponseSelector
     epochs: 100
   - name: FallbackClassifier
     threshold: 0.3
     ambiguity_threshold: 0.1
2

really? I tried to train a nlu model with rasa 2.8.1 but encountered the same error my pipeline is

  • name: SpacyNLP model: zh_core_web_sm
    • name: SpacyTokenizer
    • name: LanguageModelFeaturizer model_name: bert model_weights: bert-base-chinese
    • name: DIETClassifier epochs: 100 constrain_similarities: true