Upload 2 files

app.py

@@ -10,7 +10,6 @@ Original file is located at
 import torch
 import torch.nn as nn
 from torch.nn import functional as F
-import gradio as gr
 import requests
 
 # hyperparameters
@@ -21,13 +20,14 @@ n_head = 4
 n_layer = 4
 dropout = 0.0
 device = 'cuda' if torch.cuda.is_available() else 'cpu'
+learning_rate = 1e-3
+max_iters = 5000  # Number of training iterations
 
 # File path for saving the Book of Mormon text
 file_path = "Book of Mormon.txt"
 
-url = "https://raw.githubusercontent.com/huang-0505/LLM/refs/heads/main/Book%20of%20Mormon.txt"
-
 # Download and save the file
+url = "https://raw.githubusercontent.com/huang-0505/LLM/refs/heads/main/Book%20of%20Mormon.txt"
 response = requests.get(url)
 with open("Book of Mormon.txt", "w", encoding="utf-8") as f:
     f.write(response.text)
@@ -43,6 +43,22 @@ itos = {i: ch for i, ch in enumerate(chars)}
 encode = lambda s: [stoi[c] for c in s]
 decode = lambda l: ''.join([itos[i] for i in l])
 
+# Encode the dataset
+data = torch.tensor(encode(text), dtype=torch.long)
+
+# Split into training and validation sets
+n = int(0.9 * len(data))  # 90% training, 10% validation
+train_data = data[:n]
+val_data = data[n:]
+
+# Function to get batches of data
+def get_batch(split):
+    data = train_data if split == "train" else val_data
+    ix = torch.randint(len(data) - block_size, (batch_size,))
+    x = torch.stack([data[i:i + block_size] for i in ix])
+    y = torch.stack([data[i + 1:i + block_size + 1] for i in ix])
+    return x.to(device), y.to(device)
+
 # Model definition
 class BigramLanguageModel(nn.Module):
     def __init__(self):
@@ -127,34 +143,61 @@ class FeedForward(nn.Module):
     def forward(self, x):
         return self.net(x)
 
-# Load pre-trained model
-model = BigramLanguageModel()
-model.load_state_dict(torch.load('model.pth', map_location=device))
-model.eval()
+# Initialize model and optimizer
+model = BigramLanguageModel().to(device)
+optimizer = torch.optim.AdamW(model.parameters(), lr=learning_rate)
+
+# Training loop
+for iter in range(max_iters):
+    xb, yb = get_batch("train")
+    logits, loss = model(xb, yb)
+    optimizer.zero_grad()
+    loss.backward()
+    optimizer.step()
+
+    if iter % 100 == 0:
+        print(f"Step {iter}: Loss = {loss.item()}")
+
+# Save the model
+torch.save(model.state_dict(), "model.pth")
+print("Model trained and saved as 'model.pth'")
+
+!pip install gradio
+
+import gradio as gr
+
 
-# Gradio functions
 def ask_question(question, max_new_tokens=100):
+    # Format the input context
     context_text = f"Q: {question}\nA:"
-    context_tokens = torch.tensor(encode(context_text), dtype=torch.long, device=device).unsqueeze(0)
+    context_tokens = torch.tensor([encode(context_text)], dtype=torch.long, device=device)
+
+    # Generate the response
     generated_tokens = model.generate(context_tokens, max_new_tokens=max_new_tokens)
+
+    # Decode the generated tokens into text
     generated_text = decode(generated_tokens[0].tolist())
-    return generated_text.split("A:")[1].strip()
 
+    # Extract the answer (after "A:")
+    answer = generated_text.split("A:")[1].strip()
+    return answer
+
+# Function to process the question
 def chatbot_response(question):
-    try:
-        return ask_question(question)
-    except Exception as e:
-        return f"Error: {e}"
+  try:
+     answer = ask_question(question)
+     return f"Q: {question}\nA: {answer}"
+  except Exception as e:
+     return f"Error: {e}"
 
-# Gradio Interface
+# Create a Gradio interface
 demo = gr.Interface(
     fn=chatbot_response,
     inputs="text",
     outputs="text",
     title="Religious Chatbot",
-    description="Ask questions about the Book of Mormon."
+    description="Ask questions about the book of Mormon, and the chatbot will generate answers based on its knowledge."
 )
 
 # Launch the app
-demo.launch()
-
+demo.launch(share=True)

model.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:faf7a48948c35f21225ee5871bf989f7a14d70a4cb1d138cd21702499cb7cb0d
+oid sha256:d7a81949af5a132ffeeb6d9c6f0224663ebc79a4b64ac4254fc652b65280d478
 size 955314