Update app.py

app.py

@@ -1,69 +1,176 @@
+import os
+import warnings
+
+os.environ["TF_CPP_MIN_LOG_LEVEL"] = "3"
+warnings.filterwarnings("ignore")
+
+import torch
 import gradio as gr
-from huggingface_hub import InferenceClient
+from transformers import AutoTokenizer, AutoModelForCausalLM, logging as hf_logging
 
+hf_logging.set_verbosity_error()
 
-def respond(
-    message,
-    history: list[dict[str, str]],
-    system_message,
-    max_tokens,
-    temperature,
-    top_p,
-    hf_token: gr.OAuthToken,
-):
-    """
-    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(token=hf_token.token, model="openai/gpt-oss-20b")
+# ── Config ────────────────────────────────────────────────────────────────────
 
-    messages = [{"role": "system", "content": system_message}]
+MODEL_ID = "SupraLabs/Supra-50M-Instruct"
+DEVICE = "cpu"
 
-    messages.extend(history)
+# ── Load model ────────────────────────────────────────────────────────────────
 
-    messages.append({"role": "user", "content": message})
+print(f"[*] Loading {MODEL_ID} on CPU...")
 
-    response = ""
+tokenizer = AutoTokenizer.from_pretrained(MODEL_ID, clean_up_tokenization_spaces=False)
+model = AutoModelForCausalLM.from_pretrained(MODEL_ID, torch_dtype=torch.float32)
+model.eval()
 
-    for message in client.chat_completion(
-        messages,
-        max_tokens=max_tokens,
-        stream=True,
-        temperature=temperature,
-        top_p=top_p,
-    ):
-        choices = message.choices
-        token = ""
-        if len(choices) and choices[0].delta.content:
-            token = choices[0].delta.content
+print("[+] Model ready.")
 
-        response += token
-        yield response
+# ── Prompt builder (Alpaca format) ────────────────────────────────────────────
 
+def build_prompt(history: list[dict], system: str) -> str:
+    """Convert chat history into the Alpaca instruct format the model expects."""
+    parts = []
 
+    if system.strip():
+        parts.append(
+            "Below is an instruction that describes a task. "
+            "Write a response that appropriately completes the request.\n\n"
+            f"### Instruction:\n{system}\n\n### Response:\nUnderstood.\n"
+        )
+
+    for msg in history:
+        role, content = msg["role"], msg["content"]
+        if role == "user":
+            parts.append(
+                "Below is an instruction that describes a task. "
+                "Write a response that appropriately completes the request.\n\n"
+                f"### Instruction:\n{content}\n\n### Response:\n"
+            )
+        elif role == "assistant":
+            parts.append(content + "\n")
+
+    return "".join(parts)
+
+
+# ── Generation ────────────────────────────────────────────────────────────────
+
+def generate_response(
+    message: str,
+    history: list[dict],
+    system_prompt: str,
+    max_new_tokens: int,
+    temperature: float,
+    top_p: float,
+    repetition_penalty: float,
+) -> str:
+    history = history + [{"role": "user", "content": message}]
+    prompt = build_prompt(history, system_prompt)
+
+    inputs = tokenizer(prompt, return_tensors="pt").to(DEVICE)
+
+    with torch.no_grad():
+        output_ids = model.generate(
+            **inputs,
+            max_new_tokens=max_new_tokens,
+            do_sample=temperature > 0,
+            temperature=temperature if temperature > 0 else 1.0,
+            top_p=top_p,
+            top_k=50,
+            repetition_penalty=repetition_penalty,
+            pad_token_id=tokenizer.pad_token_id or tokenizer.eos_token_id,
+            eos_token_id=tokenizer.eos_token_id,
+        )
+
+    new_tokens = output_ids[0][inputs["input_ids"].shape[-1]:]
+    return tokenizer.decode(new_tokens, skip_special_tokens=True).strip()
+
+
+# ── UI ────────────────────────────────────────────────────────────────────────
+
+DESCRIPTION = """
+<div style="text-align:center; padding: 8px 0 4px;">
+  <h1 style="font-size:2rem; margin:0;">🦅 Supra-50M Instruct</h1>
+  <p style="color:#888; margin:4px 0 0;">50M-parameter chat model by <a href="https://huggingface.co/SupraLabs" target="_blank">SupraLabs</a> — running on CPU</p>
+</div>
 """
-For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
-"""
-chatbot = 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)",
-        ),
-    ],
-)
-
-with gr.Blocks() as demo:
-    with gr.Sidebar():
-        gr.LoginButton()
-    chatbot.render()
+
+with gr.Blocks(title="Supra-50M Instruct", theme=gr.themes.Soft()) as demo:
+    gr.HTML(DESCRIPTION)
+
+    with gr.Row():
+        with gr.Column(scale=3):
+            chatbot = gr.Chatbot(
+                label="Chat",
+                type="messages",
+                height=480,
+                show_copy_button=True,
+            )
+            with gr.Row():
+                msg_box = gr.Textbox(
+                    placeholder="Type your message…",
+                    show_label=False,
+                    scale=5,
+                    lines=1,
+                    max_lines=4,
+                    submit_btn=True,
+                )
+
+        with gr.Column(scale=1, min_width=220):
+            gr.Markdown("### ⚙️ Parameters")
+            system_prompt = gr.Textbox(
+                label="System prompt",
+                value="You are a helpful and concise assistant.",
+                lines=3,
+            )
+            max_new_tokens = gr.Slider(
+                label="Max new tokens", minimum=32, maximum=512, value=256, step=32
+            )
+            temperature = gr.Slider(
+                label="Temperature", minimum=0.1, maximum=1.5, value=0.7, step=0.05
+            )
+            top_p = gr.Slider(
+                label="Top-p", minimum=0.1, maximum=1.0, value=0.9, step=0.05
+            )
+            repetition_penalty = gr.Slider(
+                label="Repetition penalty", minimum=1.0, maximum=1.5, value=1.15, step=0.05
+            )
+            clear_btn = gr.Button("🗑️ Clear chat", variant="secondary")
+
+    # ── State & wiring ────────────────────────────────────────────────────────
+
+    chat_history = gr.State([])
+
+    def on_submit(message, history, system, max_tok, temp, top_p_val, rep_pen):
+        if not message.strip():
+            return history, history, ""
+
+        response = generate_response(
+            message, history, system, max_tok, temp, top_p_val, rep_pen
+        )
+
+        history = history + [
+            {"role": "user", "content": message},
+            {"role": "assistant", "content": response},
+        ]
+        return history, history, ""
+
+    msg_box.submit(
+        fn=on_submit,
+        inputs=[msg_box, chat_history, system_prompt, max_new_tokens, temperature, top_p, repetition_penalty],
+        outputs=[chatbot, chat_history, msg_box],
+    )
+
+    clear_btn.click(
+        fn=lambda: ([], [], ""),
+        outputs=[chatbot, chat_history, msg_box],
+    )
+
+    gr.Markdown(
+        "<p style='text-align:center; color:#aaa; font-size:0.8rem; margin-top:12px;'>"
+        "Model: <a href='https://huggingface.co/SupraLabs/Supra-50M-Instruct' target='_blank'>SupraLabs/Supra-50M-Instruct</a> — "
+        "Apache 2.0 License — © SupraLabs 2026</p>"
+    )
 
 
 if __name__ == "__main__":
-    demo.launch()
+    demo.launch()