search

google-research / albert

ALBERT: A Lite BERT for Self-supervised Learning of Language Representations
Apache License 2.0
3.24k stars 569 forks source link

ALBERT serving (`tf.contrib.resampler` should be done) #128

Closed dmoham1476 closed 4 years ago

dmoham1476 commented 4 years ago

Hi

I am trying to serve a fine-tuned ALBERT model, but facing this issue. I made sure the tf version is same for training and serving (1.15). I see that for training, in the file_based_input_fn_builder in classifier_utils uses contrib_data but not anywhere else. Is there any suggestion to fix this issue?

2020-01-13 18:30:09.440631: E tensorflow_serving/util/retrier.cc:37] Loading servable: {name: albert_intent_entity version: 1} failed: Not found: Op type not registered 'BatchMatMulV2' in binary running on 82ab054f3248. Make sure the Op and Kernel are registered in the binary running in this process. Note that if you are loading a saved graph which used ops from tf.contrib, accessing (e.g.) tf.contrib.resampler should be done before importing the graph, as contrib ops are lazily registered when the module is first accessed.

dmoham1476 commented 4 years ago

This issue went away when I trained with tensorflow 1.13