Hi, I downloaded your bert base model for MS MARCO, and try to do evaluation only. I changed the output_dir to be decompressed directory of bert base model. But it seems not initialize from the fine-tuned model properly. Any suggestion?
INFO:tensorflow:***** Running evaluation *****
I1214 12:47:57.898369 140091554572096 run_msmarco.py:404] ***** Running evaluation *****
INFO:tensorflow: Batch size = 8
I1214 12:47:57.898453 140091554572096 run_msmarco.py:405] Batch size = 8
INFO:tensorflow:Computing metrics...
I1214 12:47:57.898611 140091554572096 run_msmarco.py:418] Computing metrics...
INFO:tensorflow:Could not find trained model in model_dir: /search/odin/Data/marco-passage-ranking/models/BERT_Base_trained_on_MSMARCO/, running initialization to predict.
I1214 12:47:57.898884 140091554572096 estimator.py:615] Could not find trained model in model_dir: /search/odin/Data/marco-passage-ranking/models/BERT_Base_trained_on_MSMARCO/, running initialization to predict.
Hi, I downloaded your bert base model for MS MARCO, and try to do evaluation only. I changed the
output_dirto be decompressed directory of bert base model. But it seems not initialize from the fine-tuned model properly. Any suggestion?