Create app.py

app.py

@@ -0,0 +1,75 @@
+import gradio as gr
+import torch
+import gdown
+from torch.nn import functional as F
+
+# Download the model from Google Drive
+model_url = "https://drive.google.com/uc?id=14k2xUrvJ32trhLCzV2_O7klreBBA3dUu"
+output = "model.pth"
+gdown.download(model_url, output, quiet=False)
+
+# Load and prepare the model
+class GPTLanguageModel(nn.Module):
+    def __init__(self, vocab_size):
+        super().__init__()
+        self.token_embedding_table = nn.Embedding(vocab_size, n_embd)
+        self.position_embedding_table = nn.Embedding(block_size, n_embd)
+        self.blocks = nn.Sequential(*[Block(n_embd, n_head=n_head) for _ in range(n_layer)])
+        self.ln_f = nn.LayerNorm(n_embd)
+        self.lm_head = nn.Linear(n_embd, vocab_size)
+
+    def forward(self, index, targets=None):
+        B, T = index.shape
+        tok_emb = self.token_embedding_table(index)
+        pos_emb = self.position_embedding_table(torch.arange(T, device=device))
+        x = tok_emb + pos_emb
+        x = self.blocks(x)
+        x = self.ln_f(x)
+        logits = self.lm_head(x)
+        if targets is None:
+            loss = None
+        else:
+            B, T, C = logits.shape
+            logits = logits.view(B * T, C)
+            targets = targets.view(B * T)
+            loss = F.cross_entropy(logits, targets)
+        return logits, loss
+
+    def generate(self, index, max_new_tokens):
+        for _ in range(max_new_tokens):
+            index_cond = index[:, -block_size:]
+            logits, loss = self.forward(index_cond)
+            logits = logits[:, -1, :]
+            probs = F.softmax(logits, dim=-1)
+            index_next = torch.multinomial(probs, num_samples=1)
+            index = torch.cat((index, index_next), dim=1)
+        return index
+
+# Load the model
+model = GPTLanguageModel(vocab_size)
+model.load_state_dict(torch.load(output, map_location=device)["modelState"])
+model.eval()
+model.to(device)
+
+# Function to generate a response
+def respond(message, max_tokens=512):
+    context = torch.tensor(encode(message), dtype=torch.long, device=device).unsqueeze(0)
+    response = decode(model.generate(context, max_new_tokens=max_tokens)[0].tolist())
+    return response
+
+# Gradio interface
+iface = gr.Interface(
+    fn=respond,
+    inputs=[
+        gr.Textbox(lines=5, label="Message", value="Hi Harry Potter"),
+        gr.Slider(minimum=100, maximum=2048, value=256, label="Max Tokens"),
+    ],
+    outputs="text",
+    title="LLM Prompting App",
+    description="An LLM that generates responses based on a prompt. You can adjust the maximum tokens.",
+    theme="huggingface",
+)
+
+# Launch the interface
+if __name__ == "__main__":
+    iface.launch()