Commit all

app.py

@@ -0,0 +1,88 @@
+import gradio as gr
+import torch
+from pathlib import Path
+import sys
+
+# Add Training directory to path to import model2
+SCRIPT_DIR = Path(__file__).parent
+TRAINING_DIR = SCRIPT_DIR / "Training"
+sys.path.insert(0, str(TRAINING_DIR))
+
+from model2 import miniGPT, GPTConfig
+
+# Load model once at startup
+DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
+print(f"Loading model on device: {DEVICE}")
+
+# Use the trained checkpoint
+checkpoint_path = SCRIPT_DIR / "Training" / "trained_model_54m400k_1epoch"
+if checkpoint_path.exists():
+    print(f"Loading checkpoint from {checkpoint_path}")
+    trainer = miniGPT.load(str(checkpoint_path), map_location=DEVICE)
+    trainer = trainer.to(DEVICE)
+    trainer.model.eval()
+    print("Model loaded successfully!")
+else:
+    raise FileNotFoundError(f"Checkpoint not found at {checkpoint_path}")
+
+def generate_recipe(prompt, max_length=300, temperature=0.85, top_k=50, top_p=0.95):
+    """Generate a recipe given a prompt."""
+    with torch.no_grad():
+        generated = trainer.generate(
+            prompt,
+            max_length=max_length,
+            temperature=temperature,
+            top_k=top_k,
+            top_p=top_p,
+            repetition_penalty=1.2,
+        )
+    return generated
+
+# Gradio interface
+with gr.Blocks(title="Recipe Generator") as demo:
+    gr.Markdown("# 🍳 Recipe Generator")
+    gr.Markdown("Generate creative recipes using a 54M parameter GPT model trained on recipe data.")
+    
+    with gr.Row():
+        with gr.Column(scale=2):
+            prompt_input = gr.Textbox(
+                label="Recipe Prompt",
+                placeholder="e.g., Recipe for chocolate cake:\n\nIngredients:",
+                lines=3,
+                value="Recipe for chocolate ",
+            )
+        with gr.Column():
+            generate_btn = gr.Button("Generate Recipe", variant="primary")
+    
+    output = gr.Textbox(label="Generated Recipe", lines=10)
+    
+    with gr.Row():
+        max_length = gr.Slider(50, 500, 300, step=50, label="Max Length")
+        temperature = gr.Slider(0.1, 2.0, 0.85, step=0.1, label="Temperature")
+    
+    with gr.Row():
+        top_k = gr.Slider(0, 100, 50, step=5, label="Top-K")
+        top_p = gr.Slider(0.0, 1.0, 0.95, step=0.05, label="Top-P")
+    
+    generate_btn.click(
+        fn=generate_recipe,
+        inputs=[prompt_input, max_length, temperature, top_k, top_p],
+        outputs=output,
+    )
+    
+    # Example prompts
+    gr.Examples(
+        examples=[
+            ["Recipe for chocolate ", 300, 0.85, 50, 0.95],
+            ["Recipe for a delicious apple pie:\n\nIngredients:\n", 300, 0.85, 50, 0.95],
+            ["Ingredients: flour butter sugar", 300, 0.85, 50, 0.95],
+            ["Instructions: preheat oven to 350", 300, 0.85, 50, 0.95],
+        ],
+        inputs=[prompt_input, max_length, temperature, top_k, top_p],
+        outputs=output,
+        fn=generate_recipe,
+        cache_examples=False,
+    )
+
+if __name__ == "__main__":
+    demo.launch()

trained_model_small/model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:53b2a9f9bcd41224d5b9b1fa4ff3c15dcea18f155d8a99e6b00d8aaca2a2f154
+size 120199835

trained_model_small/tokenizer_config.json

@@ -0,0 +1,12 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": "<|endoftext|>",
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "is_local": false,
+  "model_max_length": 1024,
+  "pad_token": "<|endoftext|>",
+  "tokenizer_class": "GPT2Tokenizer",
+  "unk_token": "<|endoftext|>"
+}