Create main.py
Browse files
main.py
ADDED
@@ -0,0 +1,28 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from transformers import GPT2LMHeadModel, GPT2Tokenizer
|
2 |
+
import torch
|
3 |
+
|
4 |
+
# Load the GPT2 tokenizer and model
|
5 |
+
tokenizer = GPT2Tokenizer.from_pretrained('gpt2')
|
6 |
+
model = GPT2LMHeadModel.from_pretrained('gpt2')
|
7 |
+
|
8 |
+
# Load the training data
|
9 |
+
with open('train.txt', 'r') as f:
|
10 |
+
text = f.read()
|
11 |
+
|
12 |
+
# Tokenize the training data
|
13 |
+
input_ids = tokenizer.encode(text, return_tensors='pt')
|
14 |
+
|
15 |
+
# Train the model
|
16 |
+
optimizer = torch.optim.Adam(model.parameters(), lr=5e-5)
|
17 |
+
model.train()
|
18 |
+
for i in range(100):
|
19 |
+
outputs = model(input_ids, labels=input_ids)
|
20 |
+
loss = outputs[0]
|
21 |
+
loss.backward()
|
22 |
+
optimizer.step()
|
23 |
+
optimizer.zero_grad()
|
24 |
+
|
25 |
+
print(f'Epoch {i+1}, Loss: {loss.item()}')
|
26 |
+
|
27 |
+
# Save the trained model
|
28 |
+
model.save_pretrained('my_gpt_model')
|