app.py

# -*- coding: utf-8 -*-

import torch
from torch import nn
import torch.nn.functional as F
from transformers import AutoTokenizer, AutoModelWithLMHead
from functools import lru_cache
from tokenizers import ByteLevelBPETokenizer
from tokenizers.processors import BertProcessing

import pandas as pd


def setup_tokenizer():
  tokenizer = AutoTokenizer.from_pretrained('distilroberta-base')
  tokenizer.save_pretrained("tokenizer")


import os
os.system("mkdir -p tokenizer")
setup_tokenizer()


# from https://github.com/digantamisra98/Mish/blob/b5f006660ac0b4c46e2c6958ad0301d7f9c59651/Mish/Torch/mish.py
@torch.jit.script
def mish(input):
    return input * torch.tanh(F.softplus(input))
  
class Mish(nn.Module):
    def forward(self, input):
        return mish(input)

class EmoModel(nn.Module):
    def __init__(self, base_model, n_classes=2, base_model_output_size=768, dropout=0.05):
        super().__init__()
        self.base_model = base_model
        
        self.classifier = nn.Sequential(
            nn.Dropout(dropout),
            nn.Linear(base_model_output_size, base_model_output_size),
            Mish(),
            nn.Dropout(dropout),
            # originally, n_classes = 6
            # now, we want to use VA, change it to 2
            nn.Linear(base_model_output_size, n_classes)
        )
        
        for layer in self.classifier:
            if isinstance(layer, nn.Linear):
                layer.weight.data.normal_(mean=0.0, std=0.02)
                if layer.bias is not None:
                    layer.bias.data.zero_()

    def forward(self, input_, *args):
        X, attention_mask = input_
        hidden_states = self.base_model(X, attention_mask=attention_mask)
        return self.classifier(hidden_states[0][:, 0, :])
        

from pathlib import Path
pretrained_path = "on_plurk_new_fix_data_arch_1_epoch_2_bs_16.pt"
assert Path(pretrained_path).is_file()

model = EmoModel(AutoModelWithLMHead.from_pretrained("distilroberta-base").base_model)
model.load_state_dict(torch.load(pretrained_path,map_location=torch.device('cpu')))
model.eval()

from functools import lru_cache
@lru_cache(maxsize=1)
def get_tokenizer(max_tokens=512):
    from tokenizers import ByteLevelBPETokenizer
    from tokenizers.processors import BertProcessing
    # add error checking
    voc_file = "tokenizer/vocab.json"
    merg_file = "tokenizer/merges.txt"

    import os.path
    if not os.path.isfile(voc_file) or not os.path.isfile(merg_file):
        setup_tokenizer()

    t = ByteLevelBPETokenizer(
        voc_file,
        merg_file
    )
    t._tokenizer.post_processor = BertProcessing(
        ("</s>", t.token_to_id("</s>")),
        ("<s>", t.token_to_id("<s>")),
    )
    t.enable_truncation(max_tokens)
    t.enable_padding(length=max_tokens, pad_id=t.token_to_id("<pad>"))
    return t

# Cell
def convert_text_to_tensor(text, tokenizer=None):
    if tokenizer is None:
        tokenizer = get_tokenizer()
    enc = tokenizer.encode(text)
    X = torch.tensor(enc.ids).unsqueeze(0)
    Attn = torch.tensor(enc.attention_mask).unsqueeze(0)
    return (X, Attn)

def get_output(text, model, tokenizer=None, return_tensor=False):
    # we should add try/Except error handling for "model" argument
    # , but i consider it to be ugly
    import torch
    with torch.no_grad():
        model.eval()
        out = model(convert_text_to_tensor(text, tokenizer))
    if return_tensor == True:
      return out

    else: # return [float, float]
      # remember to make it a 1-D tensor
      tt = out[0]
      return float(tt[0]), float(tt[1])

import gradio as gr

def fn2(text, model=model, return_tensor=False):
  out = get_output(text,model, return_tensor=return_tensor)
  return out

interface = gr.Interface(
    fn = fn2,
    inputs="text",
    outputs=["number", "number"]
)

interface.launch()