Update app.py

app.py

@@ -1,125 +1,36 @@
-import os
-os.system("pip install git+https://github.com/shumingma/transformers.git accelerate")
-
-import threading
-import torch
-import torch._dynamo
-torch._dynamo.config.suppress_errors = True
-
-from transformers import (
-    AutoModelForCausalLM,
-    AutoTokenizer,
-    TextIteratorStreamer,
-)
-import gradio as gr
 import spaces
-
-model_id = "microsoft/bitnet-b1.58-2B-4T"
-
-# tokenizer unchanged
-tokenizer = AutoTokenizer.from_pretrained(model_id)
-
-# load on CPU by default; no device_map
-model = AutoModelForCausalLM.from_pretrained(
-    model_id,
-    torch_dtype=torch.bfloat16,       # CPU can emulate bfloat16
-    low_cpu_mem_usage=True            # reduces peak RAM (requires accelerate) 
-)
-
-print(next(model.parameters()).device)
-
 @spaces.GPU(duration=15)
 def gpu():
     print("[GPU] | GPU maintained.")
     
-def respond(
-    message: str,
-    history: list[tuple[str, str]],
-    system_message: str,
-    max_tokens: int,
-    temperature: float,
-    top_p: float,
-):
-    messages = [{"role": "system", "content": system_message}]
-    for user_msg, bot_msg in history:
-        if user_msg:
-            messages.append({"role": "user", "content": user_msg})
-        if bot_msg:
-            messages.append({"role": "assistant", "content": bot_msg})
-    messages.append({"role": "user", "content": message})
-
-    prompt = tokenizer.apply_chat_template(
-        messages, tokenize=False, add_generation_prompt=True
-    )
-    inputs = tokenizer(prompt, return_tensors="pt")
-
-    streamer = TextIteratorStreamer(
-        tokenizer, skip_prompt=True, skip_special_tokens=True
-    )
-    generate_kwargs = dict(
-        **inputs,
-        streamer=streamer,
-        max_new_tokens=max_tokens,
-        temperature=temperature,
-        top_p=top_p,
-        do_sample=True,
-    )
-    thread = threading.Thread(target=model.generate, kwargs=generate_kwargs)
-    thread.start()
-
-    response = ""
-    for new_text in streamer:
-        response += new_text
-        yield response
+import os
+import subprocess
+import gradio as gr
 
-demo = gr.ChatInterface(
-    fn=respond,
-    title="Bitnet-b1.58-2B-4T Chatbot",
-    description="This chat application is powered by Microsoft's SOTA Bitnet-b1.58-2B-4T and designed for natural and fast conversations.",
-    examples=[
-        [
-            "Hello! How are you?",
-            "You are a helpful AI assistant for everyday tasks.",
-            512,
-            0.7,
-            0.95,
-        ],
-        [
-            "Can you code a snake game in Python?",
-            "You are a helpful AI assistant for coding.",
-            2048,
-            0.7,
-            0.95,
-        ],
-    ],
-    additional_inputs=[
-        gr.Textbox(
-            value="You are a helpful AI assistant.",
-            label="System message"
-        ),
-        gr.Slider(
-            minimum=1,
-            maximum=8192,
-            value=2048,
-            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)"
-        ),
+MODEL_PATH = os.environ.get("MODEL_PATH", "models/BitNet-b1.58-2B-4T/ggml-model-i2_s.gguf")
+
+def generate(prompt, max_tokens=128, temperature=0.7):
+    cmd = [
+        "python", "run_inference.py",
+        "-m", MODEL_PATH,
+        "-p", prompt,
+        "-n", str(max_tokens),
+        "-temp", str(temperature)
+    ]
+    result = subprocess.run(cmd, capture_output=True, text=True)
+    return result.stdout.strip()
+
+iface = gr.Interface(
+    fn=generate,
+    inputs=[
+        gr.Textbox(lines=2, placeholder="Enter your prompt here", label="Prompt"),
+        gr.Slider(1, 512, value=128, step=1, label="Max Tokens"),
+        gr.Slider(0.0, 1.0, value=0.7, step=0.01, label="Temperature")
     ],
+    outputs=gr.Textbox(label="Completion"),
+    title="BitNet.cpp Completion Demo",
+    description="demo of bitnet.cpp inference for 1-bit llms"
 )
 
 if __name__ == "__main__":
-    demo.launch()
+    iface.launch()