Stuck on an issue?

Lightrun Answers was designed to reduce the constant googling that comes with debugging 3rd party libraries. It collects links to all the places you might be looking at while hunting down a tough bug.

And, if you’re still stuck at the end, we’re happy to hop on a call to see how we can help out.

Load and continue training NER model

See original GitHub issue

I am training a BiGRU_CRF_Model model with BERT with 1million data points and saving it.

### Load bert embeddings
bert_embed = BERTEmbedding(bert_location, task=kashgari.LABELING, sequence_length=bert_sequence_len)

### Define model
model = BiGRU_CRF_Model(bert_embed)
tf_board_callback = keras.callbacks.TensorBoard(log_dir='./logs', update_freq=1000)
eval_callback = EvalCallBack(kash_model=model, valid_x=valid_x, valid_y=valid_y, step=steps)

### Fit model
model.fit(train_x, train_y, valid_x, valid_y, batch_size=batch_size, callbacks=[eval_callback, tf_board_callback])
model.save("./BiGRU_CRF_BERT.model")

Then I am loading the model again to train it on some more data as shown below -

from kashgari.utils import load_model
model = load_model('./BiGRU_CRF_BERT.model')
model.fit(train_x, train_y, valid_x, valid_y, batch_size=batch_size, callbacks=[eval_callback, tf_board_callback])

There is no issue in model load and testing. But when I do model.fit() it gives me following error.

Traceback (most recent call last):
  File "BiLSTM_CRF_BERT.py", line 55, in <module>
    model.fit(train_x, train_y, valid_x, valid_y, epochs=epochs, batch_size=batch_size, callbacks=[eval_callback])
  File "/root/anaconda3/lib/python3.7/site-packages/kashgari/tasks/base_model.py", line 293, in fit
    **fit_kwargs)
  File "/root/anaconda3/lib/python3.7/site-packages/tensorflow/python/keras/engine/training.py", line 1433, in fit_generator
    steps_name='steps_per_epoch')
  File "/root/anaconda3/lib/python3.7/site-packages/tensorflow/python/keras/engine/training_generator.py", line 264, in model_iteration
    batch_outs = batch_function(*batch_data)
  File "/root/anaconda3/lib/python3.7/site-packages/tensorflow/python/keras/engine/training.py", line 1142, in train_on_batch
    self._assert_compile_was_called()
  File "/root/anaconda3/lib/python3.7/site-packages/tensorflow/python/keras/engine/training.py", line 2934, in _assert_compile_was_called
    raise RuntimeError('You must compile your model before '
RuntimeError: You must compile your model before training/testing. Use `model.compile(optimizer, loss)`.

How can I get this info about optimizer and loss from saved model ??

Environment

OS : Ubuntu 16.04
Python Version: Conda 3.7.3

Issue Analytics

State:
Created 4 years ago
Comments:8 (4 by maintainers)

Top GitHub Comments

3reactions

BrikerMancommented, Sep 5, 2019

layer_crf

It seems we have ignored some issue related to the CRF layer. Will fix ASAP.

0reactions

BrikerMancommented, Sep 15, 2019

This will fix in 0.5.4 version in several days. @allhelllooz

Top Results From Across the Web

Training Pipelines & Models · spaCy Usage Documentation

Training is an iterative process in which the model's predictions are compared against the reference annotations in order to estimate the gradient of...

7. How to Train spaCy NER Model

In the last notebook, we created a basic training set for a machine learning model using spaCy's EntityRuler. We were able to do...

Training Custom NER models in SpaCy to auto-detect named ...

The spaCy library allows you to train NER models by both updating an existing spacy model to suit the specific context of your...

Training Custom NER - Towards Data Science

Architecture to Train NER with Custom training data using spaCy. ... Next, load a blank model for the process to carry out the...

Training Custom NER Model Using Flair | by Akash Chauhan

The flair's documentation covers that in a brilliant way, but unlike spaCy, the data preparation for the data that is to be fed...