challenging-america-word-gap-prediction/inference.py

import torch
from torch import nn, optim
from torch.utils.data import DataLoader
import numpy as np
from collections import Counter
import string
import lzma
import pdb
import copy
from torch.utils.data import IterableDataset
import itertools
import lzma
import regex as re
import pickle
import string 
import pdb
import utils
import os
os.environ["CUDA_VISIBLE_DEVICES"] = "0"
device = 'cuda'

vocab_size = utils.vocab_size

with open("vocab.pickle", 'rb') as handle:
    vocab = pickle.load( handle)
vocab.set_default_index(vocab['<unk>'])
class Model(nn.Module):
    def __init__(self, vocab_size):
        super(Model, self).__init__()
        self.lstm_size = 150
        self.embedding_dim = 200
        self.num_layers = 1

        self.embedding = nn.Embedding(
            num_embeddings=vocab_size,
            embedding_dim=self.embedding_dim,
        )
        self.lstm = nn.LSTM(
            input_size=self.embedding_dim,
            hidden_size=self.lstm_size,
            num_layers=self.num_layers,
            batch_first=True,
            bidirectional=True,
            # dropout=0.2,
        )
        self.fc = nn.Linear(self.lstm_size*2, vocab_size)

    def forward(self, x, prev_state = None):
        embed = self.embedding(x)
        output, state = self.lstm(embed, prev_state)
        logits = self.fc(output)
        return logits, state

    def init_state(self, sequence_length):
        return (torch.zeros(self.num_layers*2, sequence_length, self.lstm_size).to(device),
                torch.zeros(self.num_layers*2, sequence_length, self.lstm_size).to(device))


model = Model(vocab_size = vocab_size).to(device)

model.load_state_dict(torch.load('lstm_step_10000.bin'))
model.eval()
def predict(model, text_splitted):
    model.eval()
    words = text_splitted

    x = torch.tensor([[vocab[w] for w in words]]).to(device)

    state_h, state_c = model.init_state(x.size()[0])

    y_pred, (state_h, state_c) = model(x, (state_h, state_c))


    last_word_logits = y_pred[0][-1]
    p = torch.nn.functional.softmax(last_word_logits, dim=0)

    top = torch.topk(p, 64)
    top_indices = top.indices.tolist()
    top_probs = top.values.tolist()
    top_words = vocab.lookup_tokens(top_indices)
    return top_words, top_probs

inference_result = []
with lzma.open(f'test-A/in.tsv.xz', 'r') as file:
    for line in file:
        line = line.decode("utf-8")
        line = line.rstrip()
        line = line.translate(str.maketrans('', '', string.punctuation))
        line_splitted_by_tab = line.split('\t')
        left_context = line_splitted_by_tab[-2]

        left_context_splitted = list(utils.get_words_from_line(left_context))

        top_words, top_probs = predict(model, left_context_splitted)

        string_to_print = ''

        sum_probs = 0
        for w, p in zip(top_words, top_probs):
            # print(top_words)
            if '<unk>' in w:
                continue
            string_to_print += f"{w}:{p} "
            sum_probs += p

        if string_to_print == '':
            inference_result.append("the:0.2 a:0.3 :0.5")
            continue
        unknow_prob = 1 - sum_probs
        string_to_print += f":{unknow_prob}"

        inference_result.append(string_to_print)

with open('test-A/out.tsv', 'w') as f:
    for line in inference_result:
        f.write(line+'\n')
print('All done')