How to use the fine-tuned bert pytorch model for classification (CoLa) task?
I do not see the argument `--do_predict`, in /examples/run_classifier.py.
However, `--do_predict` exists in the original implementation of the Bert.
The fine-tuned model is getting saving in the BERT_OUTPUT_DIR as pytorch_model.bin, but is there a simple way to reuse it through the command line?
Using Pytorch implementation from: https://github.com/huggingface/pytorch-pretrained-BERT
The command which I am using to execute the code is given below:
python run_classifier.py \ --task_name CoLA \ --do_train \ --do_eval \ --do_lower_case \ --data_dir ./split/ \ --bert_model bert-base-uncased \ --max_seq_length 128 \ --train_batch_size 32 \ --learning_rate 2e-5 \ --num_train_epochs 3.0 \
Hi,
Refer the link:
1. https://medium.com/huggingface/multi-label-text-classification-using-bert-the-mighty-transformer-69714fa3fb3d
2.https://github.com/google-research/bert
C K Gomathy, Thank you.
However, it did not answer my question.
Well to add more clarity to the question with respect to the solution given by you:
2. It link to bert by google (Which uses tensor flow for implementation). However, my question is regarding PyTorch implementation of BERT.
1. Well, to an extent the blog in the link answers the question, but it was not something which I was looking for. I want to know that if there is a way to execute run_classifier.py script for the purpose of evaluating the model on test case, as there is an absence of '--do_predict' flag in the pytorch's implementation.
Ashwin Geet D'Sa I would recommend to ask such questions on Stack exchange as many Pytorch and deep learning related questions can be found here :
https://datascience.stackexchange.com/questions/30136/pytorch-capabilities
https://datascience.stackexchange.com/questions/tagged/pytorch
- Badges
- Science topic