Mehrab Hossain
Reputation: 11
How to use a different pre-trained BERT model with bert_score
I want you to use different pretrain bert model embeddings for the bert score. How can I do that? P, R, F1 = score(cand, ref, lang="bn", model_type="distilbert-base-uncased", verbose=True) In model_type if use my pretain model then it gives a keyError.
Upvotes: 1
Views: 517
Answers (1)
kzawisto
Reputation: 76
You need to pass num_layers configuration parameter (if it is not given, the library will search for predefined defaults in utils.py file ).
bert_score.score(['Hello world'], ['Whats up'], model_type='/home/user/bart_large', num_layers=10)
Upvotes: 0
Related Questions
- Download pre-trained sentence-transformers model locally
- Fine-tuning BERT sentence transformer model
- Try to run an NLP model with an Electra instead of a BERT model
- Does BERT model and tokenizer should be trained with same data?
- Model name 'bert-base-uncased' was not found in tokenizers
- How to further pretrain a bert model using our custom data and increase the vocab size?
- Compare cosine similarity of word with BERT model
- How to get intermediate layers' output of pre-trained BERT model in HuggingFace Transformers library?
- Access the output of several layers of pretrained DistilBERT model