AutofixCodeAI / llm.py
Start-GPT's picture
Create llm.py
e7f711b verified
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer
class AutofixCodeAILLModel(AutoModelForCausalLM):
def __init__(self, *args, **kwargs):
super().__init__(*args, **kwargs)
self.decoder = AutoDecoder(self.config.decoder_hidden_size, self.config.decoder_num_layers)
@property
def decoder(self):
return self._decoder
@decoder.setter
def decoder(self, value):
self._decoder = value
class AutoDecoder(torch.nn.Module):
def __init__(self, hidden_size, num_layers):
super().__init__()
self.layers = torch.nn.ModuleList([torch.nn.TransformerEncoderLayer(d_model=hidden_size, nhead=8, dim_feedforward=hidden_size, dropout=0.1) for _ in range(num_layers)])
def forward(self, x):
for layer in self.layers:
x = layer(x)
return x
# Load the pre-trained model and tokenizer
model_name_or_path = "autofixcodeai-base"
tokenizer = AutoTokenizer.from_pretrained(model_name_or_path)
ll_model = AutofixCodeAILLModel.from_pretrained(model_name_or_path)
# Define the custom dataset class for your AutofixCodeAI model
class CodeFixDataset(torch.utils.data.Dataset):
def __init__(self, code_snippets, fix_snippets):
self.code_snippets = code_snippets
self.fix_snippets = fix_snippets
def __len__(self):
return len(self.code_snippets)
def __getitem__(self, idx):
code = self.code_snippets[idx]["code"]
fix = self.fix_snippets[idx]["fix"]
input_ids = tokenizer.encode(code, max_length=512, return_tensors="pt", truncation=True)
attention_mask = tokenizer.encode(fix, max_length=512, return_tensors="pt", truncation=True, add_special_tokens=False)
labels = torch.tensor(tokenizer.encode(fix, return_tensors="pt", add_special_tokens=False)).flatten()
return {"input_ids": input_ids, "attention_mask": attention_mask, "labels": labels}
# Load the dataset and create a data loader
dataset = CodeFixDataset(code_snippets, fix_snippets)
data_loader = torch.utils.data.DataLoader(dataset, batch_size=32, shuffle=True)
# Define the custom trainer class for your AutofixCodeAI model
class Trainer(torch.nn.Module):
def __init__(self, model, data_loader, device="cuda"):
super().__init__()
self.model = model
self.data_loader = data_loader
self.device = device
def forward(self, input_ids, attention_mask, labels):
output = self.model(input_ids=input_ids, attention_mask=attention_mask)
loss = self.loss_fn(output, labels)
return loss
@property
def loss_fn(self):
return torch.nn.CrossEntropyLoss()
# Train the model using the custom trainer class
trainer = Trainer(ll_model, data_loader, device="cuda")
for epoch in range(5):
trainer.model.train()
total_loss = 0
for batch in data_loader:
input_ids = batch["input_ids"].to(device)
attention_mask = batch["attention_mask"].to(device)
labels = batch["labels"].to(device)
loss = trainer(input_ids, attention_mask, labels).mean()
optimizer = torch.optim.Adam(trainer.model.parameters(), lr=1e-4)
optimizer.zero_grad()
loss.backward()
optimizer.step()
total_loss += loss.item()
print(f"Epoch {epoch+1}, Loss: {total_loss / len(data_loader)}")
# Evaluate the model using the custom trainer class
trainer.model.eval()
test_loss = 0
correct = 0
with torch.no_grad():
for batch in data_loader:
input_ids = batch["input_ids"].to(device)
attention_mask = batch["attention_mask"].to(device)
labels = batch["labels"].to(device)
output = trainer(input_ids, attention_mask, labels).mean()
loss = self.loss_fn(output, labels)
test_loss += loss.item()
_, predicted = torch.max(output, 1)
correct += (predicted == labels).sum().item()
accuracy = correct / len(data_loader.dataset)
print(f"Test Loss: {test_loss / len(data_loader)}, Accuracy: {accuracy:.2f}")