Use trained DIET model from RASA NLU to predict new message

Rasa Open Source
1

Hello everyone,

I’m trying to use my trained DIET model to predict a new text message, but I got a following error:

Traceback (most recent call last): File “/home/khanhdq/anaconda3/lib/python3.8/site-packages/tensorflow/python/eager/function.py”, line 2684, in _convert_inputs_to_signature flatten_inputs = nest.flatten_up_to( File “/home/khanhdq/anaconda3/lib/python3.8/site-packages/tensorflow/python/util/nest.py”, line 949, in flatten_up_to assert_shallow_structure(shallow_tree, File “/home/khanhdq/anaconda3/lib/python3.8/site-packages/tensorflow/python/util/nest.py”, line 868, in assert_shallow_structure assert_shallow_structure(shallow_branch, input_branch, File “/home/khanhdq/anaconda3/lib/python3.8/site-packages/tensorflow/python/util/nest.py”, line 852, in assert_shallow_structure raise ValueError( ValueError: The two structures don’t have the same sequence length. Input structure has length 1, while shallow structure has length 8.

During handling of the above exception, another exception occurred:

Traceback (most recent call last): File “demo_diet_prediction.py”, line 40, in print(diet._predict(message)) File “/home/khanhdq/anaconda3/lib/python3.8/site-packages/rasa/nlu/classifiers/diet_classifier.py”, line 904, in _predict return self.model.run_inference(model_data) File “/home/khanhdq/anaconda3/lib/python3.8/site-packages/rasa/utils/tensorflow/models.py”, line 314, in run_inference ] = self._rasa_predict(batch_in) File “/home/khanhdq/anaconda3/lib/python3.8/site-packages/rasa/utils/tensorflow/models.py”, line 276, in _rasa_predict outputs = tf_utils.to_numpy_or_python_type(self._tf_predict_step(batch_in)) File “/home/khanhdq/anaconda3/lib/python3.8/site-packages/tensorflow/python/eager/def_function.py”, line 780, in call result = self._call(*args, **kwds) File “/home/khanhdq/anaconda3/lib/python3.8/site-packages/tensorflow/python/eager/def_function.py”, line 814, in _call results = self._stateful_fn(*args, **kwds) File “/home/khanhdq/anaconda3/lib/python3.8/site-packages/tensorflow/python/eager/function.py”, line 2828, in call graph_function, args, kwargs = self._maybe_define_function(args, kwargs) File “/home/khanhdq/anaconda3/lib/python3.8/site-packages/tensorflow/python/eager/function.py”, line 3170, in _maybe_define_function args, kwargs = self._function_spec.canonicalize_function_inputs( File “/home/khanhdq/anaconda3/lib/python3.8/site-packages/tensorflow/python/eager/function.py”, line 2619, in canonicalize_function_inputs inputs = _convert_inputs_to_signature( File “/home/khanhdq/anaconda3/lib/python3.8/site-packages/tensorflow/python/eager/function.py”, line 2690, in _convert_inputs_to_signature raise ValueError("Structure of Python function inputs does not match " ValueError: Structure of Python function inputs does not match input_signature: inputs: ( (array([[[1.]]], dtype=float32),)) input_signature: ( [TensorSpec(shape=(None, None, 1), dtype=tf.float32, name=None), TensorSpec(shape=(None, 3), dtype=tf.int64, name=None), TensorSpec(shape=(None,), dtype=tf.float32, name=None), TensorSpec(shape=(None,), dtype=tf.int64, name=None), TensorSpec(shape=(None, 3), dtype=tf.int64, name=None), TensorSpec(shape=(None,), dtype=tf.float32, name=None), TensorSpec(shape=(None,), dtype=tf.int64, name=None), TensorSpec(shape=(None,), dtype=tf.float32, name=None)])

Here my code:

from rasa.nlu.classifiers.diet_classifier import DIETClassifier

from rasa.shared.nlu.training_data.message import Message

from typing import Any, Dict, List, Optional, Text

from rasa.shared.nlu.constants import (

TEXT,

ENTITIES,

INTENT,

INTENT_NAME_KEY,

PREDICTED_CONFIDENCE_KEY,

)

import json

def default_output_attributes() → Dict[Text, Any]:

return {

    TEXT: "",

    INTENT: {INTENT_NAME_KEY: None, PREDICTED_CONFIDENCE_KEY: 0.0},

    ENTITIES: [],

}

model_dir = ‘/models/20220620-112417/nlu’

f = open(model_dir + ‘/metadata.json’)

list_config = json.load(f)

for i in list_config[‘pipeline’]:

try:

    if i['file'].find('DIETClassifier') >= 0:

        metadata = i

        break

except BaseException:

    pass

diet = DIETClassifier().load(model_dir=model_dir, meta=metadata)

text = ‘hello’

data = default_output_attributes()

data[TEXT] = text

message = Message(data=data)

print(diet._predict(message))

Are there other ways to use the trained DIET model to predict new messages? Thanks!