README.md

metadata

license: mit
language:
  - en
tags:
  - NLP
pipeline_tag: feature-extraction

Usage

from transformers import AutoTokenizer
from model import (
    BERTContrastiveLearning_simcse,
    BERTContrastiveLearning_simcse_w,
    BERTContrastiveLearning_samp,
    BERTContrastiveLearning_samp_w,
)

str_list = data["string"].tolist()  # Your list of strings here
tokenizer = AutoTokenizer.from_pretrained("emilyalsentzer/Bio_ClinicalBERT")
tokenized_inputs = tokenizer(
    str_list, padding=True, max_length=50, truncation=True, return_tensors="pt"
)
input_ids = tokenized_inputs["input_ids"]
attention_mask = tokenized_inputs["attention_mask"]

model1 = BERTContrastiveLearning_simcse.load_from_checkpoint(ckpt1).eval()
model2 = BERTContrastiveLearning_simcse_w.load_from_checkpoint(ckpt2).eval()
model3 = BERTContrastiveLearning_samp.load_from_checkpoint(ckpt3).eval()
model4 = BERTContrastiveLearning_samp_w.load_from_checkpoint(ckpt4).eval()

cls, _ = model(input_ids, attention_mask)  # embeddings