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 = "" # 


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
)

####################################################################################################################################