Add motivational quote generator app with fine-tuned Llama 3.2 model

- Fine-tuned Llama 3.2 3B-Instruct with LoRA on motivational quotes dataset
- Simple Gradio interface for generating personalized advice
- MLX framework for efficient inference
- 55MB LoRA adapters tracked with Git LFS

🤖 Generated with Claude Code

README.md

@@ -1,14 +1,42 @@
----
-title: Motivational Quote Generator
-emoji: 💬
-colorFrom: yellow
-colorTo: purple
-sdk: gradio
-sdk_version: 5.0.1
-app_file: app.py
-pinned: false
-license: mit
-short_description: Fine tuned Llama-3.2-3B-Instruct for motivational quote
+# 🌟 Motivational Quote Generator
+
+A fine-tuned Llama 3.2 3B model that generates motivational quotes and advice. This model has been specifically trained on a curated dataset of inspirational content to provide guidance on various life topics.
+
+**🚀 Try it live: [https://huggingface.co/spaces/balazsthomay/motivational-quote-generator](https://huggingface.co/spaces/balazsthomay/motivational-quote-generator)**
+
+## 🎯 Features
+
+- **Personalized Advice**: Get motivational quotes tailored to your specific situation
+- **Multiple Topics**: Covers perseverance, leadership, success, personal growth, and more
+- **Adjustable Creativity**: Control the temperature for more or less creative responses
+- **Fast Generation**: Optimized with LoRA fine-tuning for efficient inference
+
+## 🛠️ Technical Details
+
+- **Base Model**: Llama 3.2 3B-Instruct (4-bit quantized)
+- **Fine-tuning Method**: LoRA (Low-Rank Adaptation)
+- **Training**: 2000 iterations on curated motivational quotes dataset
+- **Framework**: MLX for efficient Apple Silicon inference
+
+## 💡 Usage
+
+Simply enter a topic you'd like advice about, such as:
+- "Give me advice about perseverance"
+- "Give me advice about overcoming fear"
+- "Give me advice about leadership"
+
+The model will generate a personalized motivational response to help inspire and guide you.
+
+## 🔧 Model Configuration
+
+- **LoRA Rank**: 8
+- **Training Iterations**: 2000
+- **Max Sequence Length**: 2048 tokens
+
+## 📊 Training Data
+
+The model was trained on a carefully curated dataset of motivational quotes, with theme-based labeling using Ollama for improved contextual understanding.
+
 ---
 
-An example chatbot using [Gradio](https://gradio.app), [`huggingface_hub`](https://huggingface.co/docs/huggingface_hub/v0.22.2/en/index), and the [Hugging Face Inference API](https://huggingface.co/docs/api-inference/index).
+*Built with ❤️ using MLX and Gradio*

app.py

@@ -1,64 +1,85 @@
 import gradio as gr
-from huggingface_hub import InferenceClient
+import mlx_lm
+from mlx_lm.sample_utils import make_sampler
 
-"""
-For more information on `huggingface_hub` Inference API support, please check the docs: https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference
-"""
-client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")
-
-
-def respond(
-    message,
-    history: list[tuple[str, str]],
-    system_message,
-    max_tokens,
-    temperature,
-    top_p,
-):
-    messages = [{"role": "system", "content": system_message}]
-
-    for val in history:
-        if val[0]:
-            messages.append({"role": "user", "content": val[0]})
-        if val[1]:
-            messages.append({"role": "assistant", "content": val[1]})
-
-    messages.append({"role": "user", "content": message})
-
-    response = ""
-
-    for message in client.chat_completion(
-        messages,
-        max_tokens=max_tokens,
-        stream=True,
-        temperature=temperature,
-        top_p=top_p,
-    ):
-        token = message.choices[0].delta.content
-
-        response += token
-        yield response
-
-
-"""
-For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
-"""
-demo = gr.ChatInterface(
-    respond,
-    additional_inputs=[
-        gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
-        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
-        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
-        gr.Slider(
-            minimum=0.1,
-            maximum=1.0,
-            value=0.95,
-            step=0.05,
-            label="Top-p (nucleus sampling)",
-        ),
-    ],
+# Load the fine-tuned model
+print("Loading fine-tuned model...")
+model, tokenizer = mlx_lm.load(
+    'mlx-community/Llama-3.2-3B-Instruct-4bit',
+    adapter_path='./models/llama3.2-3b-quotes-lora-mlx'
 )
+print("✅ Model loaded successfully!")
+
+def chat_respond(message, temperature):
+    """Generate chat response"""
+    prompt = f"{message}"
+    
+    # Generate response
+    sampler = make_sampler(temp=temperature)
+    
+    try:
+        response = mlx_lm.generate(
+            model, tokenizer, 
+            prompt=prompt, 
+            max_tokens=150, 
+            sampler=sampler
+        )
+        
+        # Clean up the response (remove the original prompt)
+        if prompt in response:
+            response = response.replace(prompt, "").strip()
+        
+        return response
+    except Exception as e:
+        return f"Error: {str(e)}"
 
+# Create simple Gradio interface
+with gr.Blocks() as demo:
+    gr.Markdown("# 🤖 Motivational Quote Generator")
+    
+    with gr.Row():
+        temperature = gr.Slider(0.1, 1.5, 0.7, label="Temperature")
+    
+    with gr.Row():
+        with gr.Column():
+            prompt_input = gr.Textbox(
+                label="Your Prompt",
+                placeholder="Give me advice about courage",
+                lines=2
+            )
+            
+            generate_btn = gr.Button("Generate", variant="primary")
+            
+            response_output = gr.Textbox(
+                label="Response",
+                lines=6,
+                interactive=False
+            )
+    
+    # Examples
+    gr.Examples(
+        examples=[
+            "Give me advice about perseverance",
+            "Give me advice about courage", 
+            "Give me advice about success",
+            "Give me advice about self-discipline"
+        ],
+        inputs=prompt_input
+    )
+    
+    # Event handlers
+    generate_btn.click(
+        fn=chat_respond,
+        inputs=[prompt_input, temperature],
+        outputs=response_output
+    )
+    
+    prompt_input.submit(
+        fn=chat_respond,
+        inputs=[prompt_input, temperature],
+        outputs=response_output
+    )
 
+# Launch interface
 if __name__ == "__main__":
-    demo.launch()
+    demo.launch()

models/0000200_adapters.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9cb263cb8abb8f809dac6db832c91c12cd18ff8754653c9b4e44fc7b40f1ab42
+size 5249791

models/0000400_adapters.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f4115facf6940836839ab4113ed1b1a149f8b99a517c71190196ef895ae2daa5
+size 5249791

models/0000600_adapters.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f9c8cdefd22ff392daa65f0c66e76876753dd544cc947c84392efac2804150c1
+size 5249791

models/0000800_adapters.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:08b2f9f5045dee0d4955181d77db937855be569bb6cbd873ae1215e97c5b4167
+size 5249791

models/0001000_adapters.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d8bf13476136b7213f8dc88745056ecbeaf8f566d2322b3b4d16c277181a92da
+size 5249791

models/0001200_adapters.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3896c58045bf7047e5feb1fa4162eede7c45217253b847a9248dbbadc346f6a3
+size 5249791

models/0001400_adapters.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6bd00dc72deb30b7c76b1a3e7b8bc8ea4e4d66c50e1f562e6d4b1173312a113f
+size 5249791

models/0001600_adapters.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:37b618e666ecf7e914cdfb8289cd7e707cd3aa743fd23ea19fc120ecf9f14150
+size 5249791

models/0001800_adapters.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fd790891ea4ea06417f571945ad71c1bfa13aa494382a719cafa75989f0e0473
+size 5249791

models/0002000_adapters.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:040c43edaa3bbef61b329653e2ff7014fa9ea3faac1505f75a0a3b685365dd54
+size 5249791

models/adapter_config.json

@@ -0,0 +1,38 @@
+{
+    "adapter_path": "/Users/thomaybalazs/Projects/quotes-finetuning/models/llama3.2-3b-quotes-lora-mlx",
+    "batch_size": 2,
+    "config": null,
+    "data": "data/training/mlx_format",
+    "fine_tune_type": "lora",
+    "grad_checkpoint": true,
+    "iters": 2000,
+    "learning_rate": 5e-05,
+    "lora_parameters": {
+        "rank": 8,
+        "dropout": 0.0,
+        "scale": 20.0
+    },
+    "lr_schedule": null,
+    "mask_prompt": false,
+    "max_seq_length": 2048,
+    "model": "mlx-community/Llama-3.2-3B-Instruct-4bit",
+    "num_layers": 16,
+    "optimizer": "adam",
+    "optimizer_config": {
+        "adam": {},
+        "adamw": {},
+        "muon": {},
+        "sgd": {},
+        "adafactor": {}
+    },
+    "resume_adapter_file": null,
+    "save_every": 200,
+    "seed": 0,
+    "steps_per_eval": 100,
+    "steps_per_report": 25,
+    "test": false,
+    "test_batches": 500,
+    "train": true,
+    "val_batches": 25,
+    "wandb": null
+}

models/adapters.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:040c43edaa3bbef61b329653e2ff7014fa9ea3faac1505f75a0a3b685365dd54
+size 5249791

requirements.txt

@@ -1,5 +1,6 @@
-huggingface_hub==0.25.2
-mlx-lm
-gradio
-torch
-transformers
+mlx-lm>=0.18.0
+gradio>=4.0.0
+mlx>=0.18.0
+transformers>=4.40.0
+torch>=2.0.0
+numpy>=1.24.0