Train command gives error

[zjjj]

The command in "Training a New Model" section (see below) will give an error message No such file or directory: 'config/pytorch_model.bin'. Should we add --load_model trained_model/pytorch_model to the command, as I noticed that there's a pytorch_model.bin file in the trained_model folder downloaded from google drive?

python code/KBQA_Runner.py  \
        --train_folder  data/train_CQ \
        --dev_folder data/dev_CQ \
        --test_folder data/test_CQ \
        --KB_file data/CQ/kb_cache.json \
        --M2N_file data/CQ/m2n_cache.json \
        --QUERY_file data/CQ/query_cache.json \
        --vocab_file data/CQ/vocab.txt \
        --output_dir trained_model/CQ \
        --config config/bert_config_CQ.json \
        --gpu_id 1\
        --save_model my_model \
        --max_hop_num 2 \
        --num_train_epochs 100 \
        --do_train 1\
        --do_eval 1\
        --do_policy_gradient 1\
        --learning_rate 1e-5 \
        --train_limit_number 150 \

[lanyunshi]

Hi,

Sorry for the late reply. The file you are looking for is the pre-trained model of BERT. It exists under the trained-model folder.