from transformers import AutoAdapterModel, AutoTokenizer, TextClassificationPipeline
from huggingface_hub import Repository

# app = Flask(__name__)

#define model
tokenizer = AutoTokenizer.from_pretrained("UBC-NLP/MARBERT")

sarcasm_adapter = Repository(local_dir="sarcasm_adapter", clone_from="nehalelkaref/sarcasm_adapter")
aoc3_adapter = Repository(local_dir="aoc3_adapter", clone_from="nehalelkaref/aoc3_adapter")
aoc4_adapter = Repository(local_dir="aoc4_adapter", clone_from="nehalelkaref/aoc4_adapter")
fusion_adapter = Repository(local_dir="fusion_adapter", clone_from="nehalelkaref/region_fusion")

model = AutoAdapterModel.from_pretrained("UBC-NLP/MARBERT")

model.load_adapter("/aoc3_adapter", set_active=True, with_head=False)
model.load_adapter("/aoc4_adapter", set_active=True, with_head=False)
model.load_adapter("/sarcasm_adapter", set_active=True, with_head=False)

model.load_adapter_fusion("/fusion_adapter/aoc(3),aoc(4),sarcasm",with_head=True, set_active=True)

pipe = TextClassificationPipeline(tokenizer=tokenizer, model=model)

# @app.route('/predict', methods=['POST'])
# def predict():
#   text = request.json['inputs']

#   prediction = pipe(text)
#   labels = {"LABEL_0":"GULF", "LABEL_1":"LEVANT","LABEL_2":"EGYPT"}
#   regions = []
#   for res in prediction:
#     regions.append(labels[res['label']])

#   return jsonify({'response': regions})