Update app.py

app.py

@@ -1,63 +1,134 @@
+import os
+from threading import Thread
+from typing import Iterator
+
 import gradio as gr
-from transformers import AutoTokenizer, AutoModelForCausalLM
-import torch
 import spaces
+import torch
+from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer
+
+DESCRIPTION = """\
+# Llama 3.2 1B Instruct Finetuned on 10% of Open Hermes Dataset
+This is a demo of [`diabolic6045/open-llama-Instruct`](https://huggingface.co/diabolic6045/open-llama-Instruct).
+"""
 
-# Load the tokenizer and model
-tokenizer = AutoTokenizer.from_pretrained("diabolic6045/open-llama-Instruct")
-model = AutoModelForCausalLM.from_pretrained("diabolic6045/open-llama-Instruct")
+MAX_MAX_NEW_TOKENS = 1024
+DEFAULT_MAX_NEW_TOKENS = 512
+MAX_INPUT_TOKEN_LENGTH = int(os.getenv("MAX_INPUT_TOKEN_LENGTH", "1024"))
+
+device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
+
+model_id = "diabolic6045/open-llama-Instruct"
+tokenizer = AutoTokenizer.from_pretrained(model_id)
+model = AutoModelForCausalLM.from_pretrained(
+    model_id,
+    device_map="auto",
+    torch_dtype=torch.bfloat16,
+)
 model.eval()
-if torch.cuda.is_available():
-    model.to('cuda')
-
-@Spaces.GPU()
-def respond(
-    message,
-    history,
-    system_message,
-    max_tokens,
-    temperature,
-    top_p,
-):
-    # Build the conversation history
-    conversation = f"System: {system_message}\n"
-    for user_msg, bot_msg in history:
-        conversation += f"User: {user_msg}\nAssistant: {bot_msg}\n"
-    conversation += f"User: {message}\nAssistant:"
-
-    # Tokenize the input
-    inputs = tokenizer(conversation, return_tensors='pt', truncation=True, max_length=1024)
-    if torch.cuda.is_available():
-        inputs = {k: v.to('cuda') for k, v in inputs.items()}
-
-    # Generate the response
-    output = model.generate(
-        **inputs,
-        max_new_tokens=max_tokens,
+
+
+@spaces.GPU(duration=90)
+def generate(
+    message: str,
+    chat_history: list[tuple[str, str]],
+    max_new_tokens: int = 1024,
+    temperature: float = 0.6,
+    top_p: float = 0.9,
+    top_k: int = 50,
+    repetition_penalty: float = 1.2,
+) -> Iterator[str]:
+    conversation = []
+    for user, assistant in chat_history:
+        conversation.extend(
+            [
+                {"role": "user", "content": user},
+                {"role": "assistant", "content": assistant},
+            ]
+        )
+    conversation.append({"role": "user", "content": message})
+
+    input_ids = tokenizer.apply_chat_template(conversation, add_generation_prompt=True, return_tensors="pt")
+    if input_ids.shape[1] > MAX_INPUT_TOKEN_LENGTH:
+        input_ids = input_ids[:, -MAX_INPUT_TOKEN_LENGTH:]
+        gr.Warning(f"Trimmed input from conversation as it was longer than {MAX_INPUT_TOKEN_LENGTH} tokens.")
+    input_ids = input_ids.to(model.device)
+
+    streamer = TextIteratorStreamer(tokenizer, timeout=20.0, skip_prompt=True, skip_special_tokens=True)
+    generate_kwargs = dict(
+        {"input_ids": input_ids},
+        streamer=streamer,
+        max_new_tokens=max_new_tokens,
         do_sample=True,
-        temperature=temperature,
         top_p=top_p,
-        pad_token_id=tokenizer.eos_token_id
+        top_k=top_k,
+        temperature=temperature,
+        num_beams=1,
+        repetition_penalty=repetition_penalty,
     )
-    response = tokenizer.decode(output[0], skip_special_tokens=True)
+    t = Thread(target=model.generate, kwargs=generate_kwargs)
+    t.start()
+
+    outputs = []
+    for text in streamer:
+        outputs.append(text)
+        yield "".join(outputs)
 
-    # Extract the assistant's reply
-    response = response[len(conversation):].strip()
-    return response
 
-# Create the Gradio interface with the Ocean theme
-demo = gr.ChatInterface(
-    fn=respond,
+chat_interface = gr.ChatInterface(
+    fn=generate,
     additional_inputs=[
-        gr.Textbox(value="You are a friendly Chatbot.", label="System Message"),
-        gr.Slider(minimum=1, maximum=512, value=256, step=1, label="Max New Tokens"),
-        gr.Slider(minimum=0.1, maximum=1.0, value=0.7, step=0.05, label="Temperature"),
-        gr.Slider(minimum=0.1, maximum=1.0, value=0.9, step=0.05, label="Top-p (Nucleus Sampling)"),
+        gr.Slider(
+            label="Max new tokens",
+            minimum=1,
+            maximum=MAX_MAX_NEW_TOKENS,
+            step=1,
+            value=DEFAULT_MAX_NEW_TOKENS,
+        ),
+        gr.Slider(
+            label="Temperature",
+            minimum=0.1,
+            maximum=4.0,
+            step=0.1,
+            value=0.6,
+        ),
+        gr.Slider(
+            label="Top-p (nucleus sampling)",
+            minimum=0.05,
+            maximum=1.0,
+            step=0.05,
+            value=0.9,
+        ),
+        gr.Slider(
+            label="Top-k",
+            minimum=1,
+            maximum=1000,
+            step=1,
+            value=50,
+        ),
+        gr.Slider(
+            label="Repetition penalty",
+            minimum=1.0,
+            maximum=2.0,
+            step=0.05,
+            value=1.2,
+        ),
     ],
-    title="Open Llama Chatbot",
-    description="Chat with an AI assistant powered by the Open Llama Instruct model.",
-    theme=gr.themes.Ocean(),
+    stop_btn=None,
+    examples=[
+        ["Hello there! How are you doing?"],
+        ["Can you explain briefly to me what is the Python programming language?"],
+        ["Explain the plot of Cinderella in a sentence."],
+        ["How many hours does it take a man to eat a Helicopter?"],
+        ["Write a 100-word article on 'Benefits of Open-Source in AI research'"],
+    ],
+    cache_examples=False,
 )
 
+with gr.Blocks(css="style.css", fill_height=True, theme=gr.themes.Ocean()) as demo:
+    gr.Markdown(DESCRIPTION)
+    gr.DuplicateButton(value="Duplicate Space for private use", elem_id="duplicate-button")
+    chat_interface.render()
+
 if __name__ == "__main__":
-    demo.launch()
+    demo.queue().launch()