Deploy Feynman Explainer Gradio app (app.py)

app.py

@@ -0,0 +1,155 @@
+"""
+Feynman Explainer — Gradio Chat App
+Runs on Hugging Face Spaces (CPU free tier).
+
+Loads qwen2.5-3b-feynman-explainer on CPU with a CPU-safe dtype.
+Streams tokens for a responsive ChatGPT-like experience.
+"""
+
+import threading
+
+try:
+    import spaces  # HF Spaces ZeroGPU shim — no-op on CPU tier
+except ImportError:
+    pass
+
+import gradio as gr
+import torch
+from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer
+
+MODEL_ID = "shabul/qwen2.5-3b-feynman-explainer"
+
+SYSTEM_PROMPT = (
+    "You are a Feynman-style explainer. For every question, build intuition "
+    "from the ground up using concrete analogies and everyday language. "
+    "No jargon until it's earned. No bullet points. Pure flowing prose. "
+    "Be conversational and enthusiastic — like Feynman genuinely loved this topic."
+)
+
+TITLE = "🔬 Feynman Explainer"
+DESCRIPTION = """
+**Ask anything.** This model explains concepts the way Richard Feynman did —
+starting with a concrete analogy, building intuition from scratch, never hiding
+behind jargon.
+
+*Built by [Shabul Abdul](https://huggingface.co/shabul), Sr. Data Scientist.
+Fine-tuned on Apple M5 MacBook Pro using [MLX](https://github.com/ml-explore/mlx).*
+
+> *"You don't understand something unless you can explain it simply."* — Feynman
+
+---
+⏱️ **CPU only** — responses take 20–40 seconds. Worth the wait.
+"""
+
+EXAMPLES = [
+    ["How does gradient descent actually work?"],
+    ["What is entropy and why does it always increase?"],
+    ["What is a p-value and why do people misuse it?"],
+    ["Why does ice float on water?"],
+    ["What is the bias-variance tradeoff?"],
+    ["How does attention work in language models?"],
+    ["What is a derivative?"],
+    ["Why does compounding interest feel like magic?"],
+]
+
+print(f"Loading model: {MODEL_ID}")
+tokenizer = AutoTokenizer.from_pretrained(MODEL_ID)
+model = AutoModelForCausalLM.from_pretrained(
+    MODEL_ID,
+    torch_dtype=torch.float32,
+    low_cpu_mem_usage=True,
+)
+model.to("cpu")
+model.eval()
+print("Model loaded.")
+
+
+def respond(message: str, history: list[dict], max_new_tokens: int, temperature: float):
+    # Build messages list from history + new message
+    messages = [{"role": "system", "content": SYSTEM_PROMPT}]
+    for h in history:
+        messages.append({"role": h["role"], "content": h["content"]})
+    messages.append({"role": "user", "content": message})
+
+    input_ids = tokenizer.apply_chat_template(
+        messages,
+        tokenize=True,
+        add_generation_prompt=True,
+        return_tensors="pt",
+    )
+
+    streamer = TextIteratorStreamer(
+        tokenizer, skip_prompt=True, skip_special_tokens=True
+    )
+
+    gen_kwargs = dict(
+        input_ids=input_ids,
+        streamer=streamer,
+        max_new_tokens=max_new_tokens,
+        temperature=temperature,
+        do_sample=temperature > 0,
+        repetition_penalty=1.1,
+    )
+
+    thread = threading.Thread(target=model.generate, kwargs=gen_kwargs)
+    thread.start()
+
+    partial = ""
+    for token in streamer:
+        partial += token
+        yield partial
+
+    thread.join()
+
+
+with gr.Blocks(
+    title=TITLE,
+    theme=gr.themes.Soft(primary_hue="blue", neutral_hue="slate"),
+    css=".gradio-container { max-width: 820px !important; margin: auto; }",
+) as demo:
+
+    gr.Markdown(f"# {TITLE}\n{DESCRIPTION}")
+
+    with gr.Row():
+        with gr.Column(scale=4):
+            max_tokens = gr.Slider(
+                100, 600, value=350, step=50,
+                label="Max response length (tokens)",
+            )
+        with gr.Column(scale=4):
+            temperature = gr.Slider(
+                0.1, 1.2, value=0.75, step=0.05,
+                label="Creativity (temperature)",
+            )
+
+    chat = gr.ChatInterface(
+        fn=respond,
+        additional_inputs=[max_tokens, temperature],
+        examples=EXAMPLES,
+        cache_examples=False,
+        type="messages",
+        chatbot=gr.Chatbot(
+            height=480,
+            placeholder="<br><br><center>Ask me to explain anything — I'll make it simple.</center>",
+            show_label=False,
+        ),
+        textbox=gr.Textbox(
+            placeholder="e.g. How does a neural network learn?",
+            container=False,
+            scale=7,
+        ),
+        submit_btn="Explain →",
+        retry_btn="Try again",
+        undo_btn="Undo",
+        clear_btn="Clear chat",
+    )
+
+    gr.Markdown(
+        "---\n"
+        "🧠 Model: [`shabul/qwen2.5-3b-feynman-explainer`](https://huggingface.co/shabul/qwen2.5-3b-feynman-explainer) · "
+        "📦 Base: `Qwen/Qwen2.5-3B-Instruct` · "
+        "🍎 Trained on Apple Silicon with [mlx-lm](https://github.com/ml-explore/mlx-lm)"
+    )
+
+if __name__ == "__main__":
+    demo.launch()