app.py

import gradio as gr
import requests
import numpy as np
import torch
import torch.nn as nn
import torchvision.transforms as transform

import os 
import logging

from transformers import WEIGHTS_NAME,AdamW,AlbertConfig,AlbertTokenizer,BertConfig,BertTokenizer
from pabee.modeling_albert import AlbertForSequenceClassification
from pabee.modeling_bert import BertForSequenceClassification
from transformers import glue_output_modes as output_modes
from transformers import glue_processors as processors

import datasets
from whitebox_utils.classifier import MyClassifier   

import random
import numpy as np
import torch
import argparse

def random_seed(seed):
    random.seed(seed)
    np.random.seed(seed)
    torch.manual_seed(seed)
    torch.cuda.manual_seed_all(seed)

logger = logging.getLogger(__name__)

# TODO: dataset model tokenizer etc.
best_model_path = {
    'albert_STS-B':'./outputs/train/albert/STS-B/checkpoint-600',
}

MODEL_CLASSES = {
    "bert": (BertConfig, BertForSequenceClassification, BertTokenizer),
    "albert": (AlbertConfig, AlbertForSequenceClassification, AlbertTokenizer),
}

model_name = 'albert'
dataset = 'STS-B'
task_name = f'{dataset}'.lower()
if task_name not in processors:
    raise ValueError("Task not found: %s" % (task_name))
processor = processors[task_name]() # transformers package-preprocessor
output_mode = output_modes[task_name] # output type
label_list = processor.get_labels()
num_labels = len(label_list) 

output_dir = f'./outputs/train/{model_name}/{dataset}'
data_dir = f'./glue_data/{dataset}'

config_class, model_class, tokenizer_class = MODEL_CLASSES[model_name]
tokenizer = tokenizer_class.from_pretrained(output_dir, do_lower_case=True)
model = model_class.from_pretrained(best_model_path[f'{model_name}_{dataset}'])

exit_type='patience'
exit_value=3

classifier = MyClassifier(model,tokenizer,label_list,output_mode,exit_type,exit_value,model_name)

def greet(text,text2,exit_pos):
    text_input = [(text,text2)]
    classifier.get_prob_time(text_input,exit_position=int(exit_pos))

iface = gr.Interface(fn=greet, inputs=["text","text","number"], outputs="text")
iface.launch()