Problem with Multi-Intents Rasa 1.7.0: Can not squeeze dim[1]...

[JulianGerhard21]

Rasa version: 1.7.0

Rasa SDK version (if used & relevant): 1.7.0

Python version: 3.6.8

Operating system (windows, osx, ...): Windows 10

Issue:
After updating to Rasa 1.7.0 I observed the below error and started to investigate the issue. After a couple of hours I narrowed the problem down to intents like:

## intent:greet+ask_weather

in combination with CountVectorsFeaturizer meaning that the problem only happens when using both in the same setup. I first thought that it had something to do with the transformers I sometimes use, but even with de_core_news_sm models it doesnt work.

I am not quite sure, if this is in any case intended and simply slipped through my attention so I wanted to bring it up.

Error (including full traceback):

2020-01-30 16:57:43 INFO     rasa.nlu.utils.spacy_utils  - Trying to load spacy model with name 'de_core_news_sm'
2020-01-30 16:57:44 INFO     transformers.file_utils  - PyTorch version 1.3.0+cpu available.
2020-01-30 16:57:44 INFO     transformers.modeling_xlnet  - Better speed can be achieved with apex installed from https://www.github.com/nvidia/apex .
2020-01-30 16:57:45 INFO     rasa.nlu.components  - Added 'SpacyNLP' to component cache. Key 'SpacyNLP-de_core_news_sm'.
2020-01-30 16:57:45 INFO     rasa.nlu.utils.spacy_utils  - Trying to load spacy model with name 'de_core_news_md'
2020-01-30 16:58:26 INFO     rasa.nlu.components  - Added 'SpacyNLP' to component cache. Key 'SpacyNLP-de_core_news_md'.
2020-01-30 16:58:26 INFO     rasa.nlu.training_data.training_data  - Training data stats:
        - intent examples: 347 (18 distinct intents)
        - Found intents: 'repeat_last_question', 'functions', 'deny+utter_mobilenumber', 'deny', 'utter_mobilenumber', 'confirm', 'reset_dialog', 'ask_outofscope', 'deny_appointment_suggestion', 'ask_repeat_contact', 'make_appointment', 'not_now', 'greet', 'goodbye', 'thankyou', 'ask_persisted_user_data', 'persist_purpose', 'select_appointment_slot'
        - Number of response examples: 0 (0 distinct response)
        - entity examples: 104 (7 distinct entities)
        - found entities: 'appointment_type', 'daytime', 'not_now_reason', 'slot_shift', 'mobilenumber', 'time', 'employee'

2020-01-30 16:58:26 INFO     rasa.nlu.model  - Starting to train component SpacyNLP
2020-01-30 16:58:26 INFO     rasa.nlu.model  - Finished training component.
2020-01-30 16:58:26 INFO     rasa.nlu.model  - Starting to train component SpacyTokenizer
2020-01-30 16:58:26 INFO     rasa.nlu.model  - Finished training component.
2020-01-30 16:58:26 INFO     rasa.nlu.model  - Starting to train component CountVectorsFeaturizer
2020-01-30 16:58:26 INFO     rasa.nlu.model  - Finished training component.
2020-01-30 16:58:26 INFO     rasa.nlu.model  - Starting to train component CountVectorsFeaturizer
2020-01-30 16:58:27 INFO     rasa.nlu.model  - Finished training component.
2020-01-30 16:58:27 INFO     rasa.nlu.model  - Starting to train component SpacyFeaturizer
2020-01-30 16:58:27 INFO     rasa.nlu.model  - Finished training component.
2020-01-30 16:58:27 INFO     rasa.nlu.model  - Starting to train component EmbeddingIntentClassifier
Tensor("mul:0", shape=(?, ?, 192), dtype=float32)
Tensor("mul_1:0", shape=(?, ?, 20), dtype=float32)
Tensor("mul_2:0", shape=(18, 2, 20), dtype=float32)
Traceback (most recent call last):
  File "c:\users\\envs\eva_env\lib\site-packages\tensorflow_core\python\framework\ops.py", line 1607, in _create_c_op
    c_op = c_api.TF_FinishOperation(op_desc)
tensorflow.python.framework.errors_impl.InvalidArgumentError: Can not squeeze dim[1], expected a dimension of 1, got 2 for 'Squeeze_2' (op: 'Squeeze') with input shapes: [18,2,20].

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "c:\users\\appdata\local\programs\python\python36\Lib\runpy.py", line 193, in _run_module_as_main
    "__main__", mod_spec)
  File "c:\users\\appdata\local\programs\python\python36\Lib\runpy.py", line 85, in _run_code
    exec(code, run_globals)
  File "C:\Users\\Envs\eva_env\Scripts\rasa.exe\__main__.py", line 7, in <module>
  File "c:\users\\envs\eva_env\lib\site-packages\rasa\__main__.py", line 76, in main
    cmdline_arguments.func(cmdline_arguments)
  File "c:\users\\envs\eva_env\lib\site-packages\rasa\cli\train.py", line 76, in train
    additional_arguments=extract_additional_arguments(args),
  File "c:\users\\envs\eva_env\lib\site-packages\rasa\train.py", line 50, in train
    additional_arguments=additional_arguments,
  File "c:\users\\appdata\local\programs\python\python36\Lib\asyncio\base_events.py", line 484, in run_until_complete
    return future.result()
  File "c:\users\\envs\eva_env\lib\site-packages\rasa\train.py", line 101, in train_async
    additional_arguments,
  File "c:\users\\envs\eva_env\lib\site-packages\rasa\train.py", line 188, in _train_async_internal
    additional_arguments=additional_arguments,
  File "c:\users\\envs\eva_env\lib\site-packages\rasa\train.py", line 245, in _do_training
    persist_nlu_training_data=persist_nlu_training_data,
  File "c:\users\\envs\eva_env\lib\site-packages\rasa\train.py", line 474, in _train_nlu_with_validated_data
    persist_nlu_training_data=persist_nlu_training_data,
  File "c:\users\\envs\eva_env\lib\site-packages\rasa\nlu\train.py", line 86, in train
    interpreter = trainer.train(training_data, **kwargs)
  File "c:\users\\envs\eva_env\lib\site-packages\rasa\nlu\model.py", line 191, in train
    updates = component.train(working_data, self.config, **context)
  File "c:\users\\envs\eva_env\lib\site-packages\rasa\nlu\classifiers\embedding_intent_classifier.py", line 769, in train
    loss, acc = self._build_tf_train_graph(session_data)
  File "c:\users\\envs\eva_env\lib\site-packages\rasa\nlu\classifiers\embedding_intent_classifier.py", line 571, in _build_tf_train_graph
    label_data["label_features"], label_data["label_mask"][0], "label"
  File "c:\users\\envs\eva_env\lib\site-packages\rasa\nlu\classifiers\embedding_intent_classifier.py", line 547, in _combine_sparse_dense_features
    output = tf.squeeze(output, axis=1)
  File "c:\users\\envs\eva_env\lib\site-packages\tensorflow_core\python\util\dispatch.py", line 180, in wrapper
    return target(*args, **kwargs)
  File "c:\users\\envs\eva_env\lib\site-packages\tensorflow_core\python\util\deprecation.py", line 507, in new_func
    return func(*args, **kwargs)
  File "c:\users\\envs\eva_env\lib\site-packages\tensorflow_core\python\ops\array_ops.py", line 3651, in squeeze
    return gen_array_ops.squeeze(input, axis, name)
  File "c:\users\\envs\eva_env\lib\site-packages\tensorflow_core\python\ops\gen_array_ops.py", line 10068, in squeeze
    "Squeeze", input=input, squeeze_dims=axis, name=name)
  File "c:\users\\envs\eva_env\lib\site-packages\tensorflow_core\python\framework\op_def_library.py", line 794, in _apply_op_helper
    op_def=op_def)
  File "c:\users\\envs\eva_env\lib\site-packages\tensorflow_core\python\util\deprecation.py", line 507, in new_func
    return func(*args, **kwargs)
  File "c:\users\\envs\eva_env\lib\site-packages\tensorflow_core\python\framework\ops.py", line 3357, in create_op
    attrs, op_def, compute_device)
  File "c:\users\\envs\eva_env\lib\site-packages\tensorflow_core\python\framework\ops.py", line 3426, in _create_op_internal
    op_def=op_def)
  File "c:\users\\envs\eva_env\lib\site-packages\tensorflow_core\python\framework\ops.py", line 1770, in __init__
    control_input_ops)
  File "c:\users\\envs\eva_env\lib\site-packages\tensorflow_core\python\framework\ops.py", line 1610, in _create_c_op
    raise ValueError(str(e))
ValueError: Can not squeeze dim[1], expected a dimension of 1, got 2 for 'Squeeze_2' (op: 'Squeeze') with input shapes: [18,2,20].

Content of configuration file (config.yml) (if relevant):

language: de
pipeline:
 - name: SpacyNLP
   case_sensitive: 1
   model: de_trf_bertbasecased_lg_smo_enhanced
 - name: SpacyTokenizer
   intent_tokenization_flag: true
   intent_split_symbol: "+"
 - name: CountVectorsFeaturizer
   strip_accents: "unicode"
 - name: CountVectorsFeaturizer
   strip_accents: "unicode"
   analyzer: "char_wb"
   min_ngram: 2
   max_ngram: 15
 - name: SpacyFeaturizer
 - name: "EmbeddingIntentClassifier"
 - name: SpacyNLP
   case_sensitive: 1
   model: de_core_news_md
 - name: SpacyTokenizer
 - name: SpacyFeaturizer
 - name: CRFEntityExtractor
 - name: EntitySynonymMapper
 - name: DucklingHTTPExtractor
   dimensions: ['time', 'duration', 'email']
   locale: de_DE
   timezone: Europe/Berlin
   url: http://localhost:8001
 - name: SpacyEntityExtractor
   dimensions: ['PER', 'LOC', 'CARDINAL']
policies:
 - name: EmbeddingPolicy
   batch_size: 50
   epochs: 500
   featurizer:
    - name: MaxHistoryTrackerFeaturizer
      max_history: 3
      state_featurizer:
       - name: LabelTokenizerSingleStateFeaturizer
 - name: FallbackPolicy
   nlu_threshold: 0.45
   ambiguity_threshold: 0.1
   core_threshold: 0.3
   fallback_action_name: 'utter_default_fallback'
 - name: MemoizationPolicy
   max_history: 5
 - name: FormPolicy
 - name: MappingPolicy

[sara-tagger]

Thanks for raising this issue, @tmbo will get back to you about it soon✨

Please also check out the docs and the forum in case your issue was raised there too 🤗

[tmbo]

@tabergma any clue where this might be coming from?

[tabergma]

Could reproduce the problem, it is definitely a bug. Thanks for reporting it 👍Will take a closer look and work on a fix.

[tabergma]

closed via #5184