Unable to train ELMO after loading all training files.

[Dragon615]

Hi I'm trying to train ELMO using my corpus.
The training script stop after loading loading all training data.
the training script stopped after getting the below message

WARNING : Error encountered when serializing lstm_output_embeddings.
Type is unsupported, or the types of the items don't match field type in CollectionDef.
'list' object has no attribute 'name'

I have divided my training corpus into 10 files each contains 3890 sentences.
I'm using 3 gpus.

The command I'm using is the following:

$python bin/train_elmo.py --train_prefix='/path to data file/*' --vocab_file '/path to vocab file' --save_dir '/path to check point/'

Can you please help

Thanks in advance!

[lvyufeng]

maybe you should use tf1.2 instead of newer version

[pzhang84]

@Dragon615 Did you figure it out? I think this warning is okay, you should run dump_weights for serializing.