import transformers import os import torch import requests MAX_LEN = 150 #256 TRAIN_BATCH_SIZE = 8 VALID_BATCH_SIZE = 4 EPOCHS = 5 # Folder to contain all the datasets from huggingface_hub import hf_hub_download DATASET_LOCATION = "" # print("hi") MODEL_PATH = hf_hub_download(repo_id="thak123/bert-emoji-latvian-twitter-classifier", filename="model.bin") # from huggingface_hub import snapshot_download # snapshot_download(repo_id="thak123/bert-emoji-latvian-twitter-classifier", allow_patterns="*.bin") device = torch.device('cuda' if torch.cuda.is_available() else 'cpu') # 7 EPOCH Version BERT_PATH = "FFZG-cleopatra/bert-emoji-latvian-twitter" # TODO check if lower casing is required # BertTokenizer TOKENIZER = transformers.BertTokenizer.from_pretrained( BERT_PATH, do_lower_case=True ) ####################################################################################################################################