# PyTorch Imports import torch import torch.nn as nn import torch.optim as optim from torch.utils.data import Dataset, DataLoader # scikit-learn Imports # from sklearn.metrics import accuracy_score, confusion_matrix from sklearn.model_selection import train_test_split # Bert imports from transformers import BertForSequenceClassification, AutoTokenizer #Default imports (pandas, numpy, matplotlib, etc.) import pandas as pd import numpy as np import matplotlib.pyplot as plt ## Select Device if torch.cuda.is_available(): DEVICE = torch.device("cuda") else: DEVICE = torch.device("cpu") class SimpleHumorDataset(Dataset): def __init__(self,tokenizer:AutoTokenizer,dataframe:pd.DataFrame,max_length:int=128): super(SimpleHumorDataset,self).__init__() self.tokenizer = tokenizer self.max_length = max_length self.text = dataframe['text'].to_list() self.labels = dataframe['is_humor'].to_list() def __getitem__(self,idx:int): text = self.text[idx] labels = self.labels[idx] encoding = self.tokenizer( text, padding="max_length", return_attention_mask = True, max_length=self.max_length, truncation = True, return_tensors = 'pt' ) input_ids = encoding['input_ids'].flatten() attention_mask = encoding['attention_mask'].flatten() return { 'input_ids': torch.as_tensor(input_ids,dtype=torch.long), 'attention_mask':torch.as_tensor(attention_mask,dtype=torch.long), 'labels':torch.tensor(labels,dtype=torch.long) } def __len__(self): return len(self.labels) class CustomBert(nn.Module): def __init__(self): super().__init__() #Bert + Custom Layers (Not a tuple any longer -- idk why) self.bfsc = BertForSequenceClassification.from_pretrained("bert-base-uncased") self.classifier = nn.Linear(2,2) self.sm = nn.Softmax(dim=1) def forward(self, input_ids, attention_mask): seq_out = self.bfsc(input_ids, attention_mask = attention_mask) x = self.classifier(seq_out.logits) return self.sm(x) def training_loop(model:CustomBert,criterion:nn.CrossEntropyLoss,optimizer:optim.AdamW,train_loader:DataLoader): model.train() total_loss = 0 for train_batch in train_loader: # Set Gradient to Zero optimizer.zero_grad() # Unpack batch values and "push" it to GPU input_ids, att_mask, labels = train_batch.values() input_ids, att_mask, labels = input_ids.to(DEVICE),att_mask.to(DEVICE),labels.to(DEVICE) # Feed Model with Data outputs = model(input_ids, attention_mask=att_mask) loss = criterion(outputs,labels) loss.backward() optimizer.step() total_loss+=loss.item() print(f"Total Loss is {(total_loss/len(train_loader)):.4f}") return (total_loss/len(train_loader)) def eval_loop(model:CustomBert,criterion:nn.CrossEntropyLoss,validation_loader:DataLoader): model.eval() total, correct = 0.0, 0.0 total_loss = 0.0 with torch.no_grad(): for val_batch in validation_loader: input_ids, att_mask ,labels = val_batch.values() input_ids, att_mask, labels = input_ids.to(DEVICE),att_mask.to(DEVICE), labels.to(DEVICE) outputs = model(input_ids,attention_mask=att_mask) loss = criterion(outputs,labels) total_loss += loss.item() predictions = torch.argmax(outputs,1) total += labels.size(0) correct += (predictions == labels).sum().item() print(f"Total Loss: {total_loss/len(validation_loader)} ### Test Accuracy {correct/total*100}%") return total_loss/len(validation_loader) if __name__ == "__main__": torch.manual_seed(501) # HYPERPARAMETERS # Set Max Epoch Amount EPOCH = 1 # DROPOUT-PROBABILITY DROPOUT = 0.1 # BATCHSIZE BATCH_SIZE = 8 #LEARNING RATE LEARNING_RATE = 1e-5 # Initialize Bert Model with dropout probability and Num End Layers mybert = CustomBert() print("Bert Initialized") mybert.to(DEVICE) # Read Raw Data from csv and save as DataFrame df = pd.read_csv("./data/hack.csv",encoding="latin1") print("Raw Data read") # Initialize BertTokenizer from Pretrained tokenizer = AutoTokenizer.from_pretrained("google-bert/bert-base-uncased",do_lower_case=True) print("Tokenizer Initialized") # print(tokenizer(df['text'][0],padding=True,truncation=True,max_length=256)) #Split DataFrame into Train and Test Sets train,test = train_test_split(df,random_state=501,test_size=.2) print("Splitted Data in Train and Test Sets") # val = [] # Create Custom Datasets for Train and Test train_data = SimpleHumorDataset(tokenizer,train) # val_data = SimpleHumorDataset(tokenizer,val) test_data = SimpleHumorDataset(tokenizer,test) print("Custom Datasets created") # Initialize Dataloader with Train and Test Sets train_loader = DataLoader(dataset=train_data,batch_size=BATCH_SIZE,shuffle=True) # val_loader = DataLoader(dataset=val_data,batch_size=BATCH_SIZE,shuffle=True) test_loader = DataLoader(dataset=test_data,batch_size=BATCH_SIZE,shuffle=False) print("DataLoaders created") # Set criterion to Cross Entropy and define Adam Optimizer with model parameters and learning rate criterion_cross_entropy = nn.CrossEntropyLoss() optimizer_adamW = optim.Adam(mybert.parameters(), lr = LEARNING_RATE) import time # Set Scheduler for dynamically Learning Rate adjustment loss_values = np.zeros(EPOCH) eval_values = np.zeros(EPOCH) start = time.time() for epoch in range(EPOCH): print(f"For {epoch+1} the Scores are: ") loss_values[epoch] = training_loop(mybert,optimizer=optimizer_adamW,criterion=criterion_cross_entropy,train_loader=train_loader) eval_values[epoch] = eval_loop(mybert,criterion=criterion_cross_entropy,validation_loader=test_loader) end = time.time() print((end-start),"seconds per epoch needed") # Visualize Training Loss plt.plot(loss_values) plt.plot(eval_values) plt.hlines(np.mean(loss_values),xmin=0,xmax=EPOCH,colors='red',linestyles="dotted",label="Average Loss") plt.hlines(np.mean(eval_values),xmin=0,xmax=EPOCH,colors='green',linestyles="dashed",label="Average Val Loss") plt.title("Test Loss") plt.xlabel("Num Epochs") plt.ylabel("Total Loss of Epoch") plt.show()