Supprimer 'main.py'
This commit is contained in:
parent
db8d6f328c
commit
fab77c4a08
83
main.py
83
main.py
|
@ -1,83 +0,0 @@
|
|||
import torch
|
||||
import torch.nn as nn
|
||||
import torch.optim as optim
|
||||
from torchtext import data
|
||||
from gensim.corpora import WikiCorpus
|
||||
from transformers import GPT2Tokenizer, GPT2Model
|
||||
from functions import *
|
||||
|
||||
# Define the hyperparameters
|
||||
num_layers = 2
|
||||
batch_size = 32
|
||||
hidden_dim = 256
|
||||
|
||||
# Load the GPT2 model
|
||||
tokenizer = GPT2Tokenizer.from_pretrained('gpt2')
|
||||
model = GPT2Model.from_pretrained('gpt2')
|
||||
|
||||
# Load the data
|
||||
wiki_corpus = WikiCorpus('enwiki-latest-pages-articles.xml.bz2')
|
||||
stackoverflow_corpus = data.TabularDataset('stackoverflow.csv', format='csv', fields=['text'])
|
||||
|
||||
# Preprocess the data
|
||||
wiki_data = [text for text in wiki_corpus]
|
||||
stackoverflow_data = [text for text in stackoverflow_corpus]
|
||||
|
||||
# Convert the data to a format compatible with PyTorch
|
||||
wiki_data = torch.tensor(wiki_data)
|
||||
stackoverflow_data = torch.tensor(stackoverflow_data)
|
||||
|
||||
# Define the Adam optimizer
|
||||
optimizer = optim.Adam(model.parameters(), lr=0.001)
|
||||
|
||||
# Define the loss function
|
||||
criterion = nn.CrossEntropyLoss()
|
||||
|
||||
# Train the model
|
||||
num_epochs=10
|
||||
labels = torch.tensor([0, 1, 1, 0, 0, 1, 0, 1, 0, 1])
|
||||
|
||||
def adjust_learning_rate(optimizer, epoch):
|
||||
"""Sets the learning rate to the initial LR decayed by 10 every 30 epochs"""
|
||||
lr = 0.001 * (0.1 ** (epoch // 30))
|
||||
for param_group in optimizer.param_groups:
|
||||
param_group['lr'] = lr
|
||||
|
||||
for epoch in range(num_epochs):
|
||||
# Forward pass
|
||||
outputs = model(wiki_data, stackoverflow_data)
|
||||
# Calculate the loss
|
||||
loss = criterion(outputs, labels)
|
||||
# Backward pass
|
||||
loss.backward()
|
||||
# Update the parameters
|
||||
optimizer.step()
|
||||
# Reset the gradients
|
||||
optimizer.zero_grad()
|
||||
# Evaluate the model
|
||||
accuracy = evaluate(model, wiki_data)
|
||||
# Save the model weights and states
|
||||
torch.save(model.state_dict(), 'model.pth')
|
||||
# Adjust the learning rate
|
||||
adjust_learning_rate(optimizer, epoch)
|
||||
|
||||
|
||||
# Define the model
|
||||
class GPT(nn.Module):
|
||||
def __init__(self, vocab_size, embedding_dim, hidden_dim, num_layers):
|
||||
super().__init__()
|
||||
self.embedding = nn.Embedding(vocab_size, embedding_dim)
|
||||
self.lstm = nn.LSTM(embedding_dim, hidden_dim, num_layers, batch_first=True)
|
||||
self.fc = nn.Linear(hidden_dim, vocab_size)
|
||||
self.gpt2 = model
|
||||
|
||||
def forward(self, x):
|
||||
# Embed the input
|
||||
x = self.embedding(x)
|
||||
# Pass through the GPT2 model
|
||||
x = self.gpt2(x)
|
||||
# Pass through the LSTM
|
||||
x, _ = self.lstm(x)
|
||||
# Pass through the fully connected layer
|
||||
x = self.fc(x)
|
||||
return x
|
Loading…
Reference in New Issue