search

ZhiGroup / Med-BERT

Med-BERT, contextualized embedding model for structured EHR data
Apache License 2.0
240 stars 60 forks source link

Pre-training issues #12

Open lishuo-0 opened 1 year ago

lishuo-0 commented 1 year ago

Hi, I ran into some problems reproducing the pre-training with my own data. An error occurred during the training loop: Error recorded from training_loop: {{function_node __inference_tf_data_experimental_map_andbatch_91357}} Key: segment_ids. can't parse serialized Example. However I printed example before training and they seemed to be fine, and the first checkpoint has been saved. May I ask why this is a problem? Do you have any suggestions? Thank you very much!

minotskaya commented 1 year ago

Hi, I have the same problem(( Any ideas how to solve this?

Ofir408 commented 1 year ago

I solved it by reducing the max_seq_length