Update app.py

app.py

@@ -1,116 +1,154 @@
 import os
-import logging
+import time
+import spaces
+import torch
+from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer
 import gradio as gr
-from huggingface_hub import InferenceClient
-from logging.handlers import RotatingFileHandler
-
-# Logging setup
-log_file = 'app_debug.log'
-logger = logging.getLogger(__name__)
-logger.setLevel(logging.DEBUG)
-file_handler = RotatingFileHandler(log_file, maxBytes=10*1024*1024, backupCount=5)
-file_handler.setFormatter(logging.Formatter('%(asctime)s - %(levelname)s - %(message)s'))
-logger.addHandler(file_handler)
-logger.debug("Application started")
-
-# Inference client setup
-client = InferenceClient("mistralai/Mistral-7B-Instruct-v0.3")
-
-def format_prompt(message, history):
-    prompt = "<s>"
-    for user_prompt, bot_response in history:
-        prompt += f"[INST] {user_prompt} [/INST]"
-        prompt += f" {bot_response}</s> "
-    prompt += f"[INST] {message} [/INST]"
-    return prompt
-
-def generate(
-    prompt, history, temperature=0.9, max_new_tokens=256, top_p=0.95, repetition_penalty=1.0,
+from threading import Thread
+
+MODEL_LIST = ["mistralai/Mistral-Nemo-Instruct-2407"]
+HF_TOKEN = os.environ.get("HF_TOKEN", None)
+MODEL = os.environ.get("MODEL_ID")
+
+TITLE = "<h1><center>Mistral-Nemo</center></h1>"
+
+PLACEHOLDER = """
+<center>
+<p>The Mistral-Nemo is a pretrained generative text model of 12B parameters trained jointly by Mistral AI and NVIDIA.</p>
+</center>
+"""
+
+
+CSS = """
+.duplicate-button {
+    margin: auto !important;
+    color: white !important;
+    background: black !important;
+    border-radius: 100vh !important;
+}
+h3 {
+    text-align: center;
+}
+"""
+
+device = "cuda" # for GPU usage or "cpu" for CPU usage
+
+tokenizer = AutoTokenizer.from_pretrained(MODEL)
+model = AutoModelForCausalLM.from_pretrained(
+    MODEL,
+    torch_dtype=torch.bfloat16,
+    device_map="auto",
+    ignore_mismatched_sizes=True)
+
+@spaces.GPU()
+def stream_chat(
+    message: str, 
+    history: list, 
+    temperature: float = 0.3, 
+    max_new_tokens: int = 1024, 
+    top_p: float = 1.0, 
+    top_k: int = 20, 
+    penalty: float = 1.2,
 ):
-    logger.debug(f"Generating response for prompt: {prompt} with history: {history}")
-    temperature = float(temperature)
-    if temperature < 1e-2:
-        temperature = 1e-2
-    top_p = float(top_p)
+    print(f'message: {message}')
+    print(f'history: {history}')
 
+    conversation = []
+    for prompt, answer in history:
+        conversation.extend([
+            {"role": "user", "content": prompt}, 
+            {"role": "assistant", "content": answer},
+        ])
+
+    conversation.append({"role": "user", "content": message})
+
+    input_text=tokenizer.apply_chat_template(conversation, tokenize=False)
+    inputs = tokenizer.encode(input_text, return_tensors="pt").to(device)
+    streamer = TextIteratorStreamer(tokenizer, timeout=60.0, skip_prompt=True, skip_special_tokens=True)
+    
     generate_kwargs = dict(
-        temperature=temperature,
-        max_new_tokens=max_new_tokens,
-        top_p=top_p,
-        repetition_penalty=repetition_penalty,
-        do_sample=True,
-        seed=42,
+        input_ids=inputs, 
+        max_new_tokens = max_new_tokens,
+        do_sample = False if temperature == 0 else True,
+        top_p = top_p,
+        top_k = top_k,
+        temperature = temperature,
+        streamer=streamer,
+        repetition_penalty=penalty,
+        pad_token_id = 10,
     )
 
-    formatted_prompt = format_prompt(prompt, history)
-
-    try:
-        stream = client.text_generation(formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=False)
-        output = ""
-
-        for response in stream:
-            output += response.token.text
-            yield output
-        logger.debug(f"Generated response: {output}")
-        return output
-    except Exception as e:
-        logger.exception("Error during text generation")
-        return "An error occurred during response generation."
-
-def update_history(history, user_input, response):
-    history.append((user_input, response))
-    return history
-
-def chat_interface(user_input, history=[], temperature=0.9, max_new_tokens=256, top_p=0.95, repetition_penalty=1.0):
-    logger.debug(f"User input: {user_input}")
-    response = generate(user_input, history, temperature, max_new_tokens, top_p, repetition_penalty)
-    history = update_history(history, user_input, response)
-    logger.debug(f"Updated history: {history}")
-    return response, history
-
-gr.ChatInterface(
-    fn=chat_interface,
-    chatbot=gr.Chatbot(show_label=False, show_share_button=False, show_copy_button=False, likeable=False, layout="panel"),
-    additional_inputs=[
-        gr.Slider(
-            label="Temperature",
-            value=0.9,
-            minimum=0.0,
-            maximum=1.0,
-            step=0.05,
-            interactive=True,
-            info="Higher values produce more diverse outputs",
-        ),
-        gr.Slider(
-            label="Max new tokens",
-            value=256,
-            minimum=0,
-            maximum=1048,
-            step=64,
-            interactive=True,
-            info="The maximum numbers of new tokens",
-        ),
-        gr.Slider(
-            label="Top-p (nucleus sampling)",
-            value=0.90,
-            minimum=0.0,
-            maximum=1,
-            step=0.05,
-            interactive=True,
-            info="Higher values sample more low-probability tokens",
-        ),
-        gr.Slider(
-            label="Repetition penalty",
-            value=1.2,
-            minimum=1.0,
-            maximum=2.0,
-            step=0.05,
-            interactive=True,
-            info="Penalize repeated tokens",
-        )
-    ],
-    title="Mistral 7B v0.3",
-    description=None
-).launch(show_api=True)
-
-logger.debug("Chat interface initialized and launched")
+    with torch.no_grad():
+        thread = Thread(target=model.generate, kwargs=generate_kwargs)
+        thread.start()
+        
+    buffer = ""
+    for new_text in streamer:
+        buffer += new_text
+        yield buffer
+
+            
+chatbot = gr.Chatbot(height=600, placeholder=PLACEHOLDER)
+
+with gr.Blocks(css=CSS, theme="Nymbo/Nymbo_Theme") as demo:
+    gr.HTML(TITLE)
+    gr.DuplicateButton(value="Duplicate Space for private use", elem_classes="duplicate-button")
+    gr.ChatInterface(
+        fn=stream_chat,
+        chatbot=chatbot,
+        fill_height=True,
+        additional_inputs_accordion=gr.Accordion(label="⚙️ Parameters", open=False, render=False),
+        additional_inputs=[
+            gr.Slider(
+                minimum=0,
+                maximum=1,
+                step=0.1,
+                value=0.3,
+                label="Temperature",
+                render=False,
+            ),
+            gr.Slider(
+                minimum=128,
+                maximum=8192,
+                step=1,
+                value=1024,
+                label="Max new tokens",
+                render=False,
+            ),
+            gr.Slider(
+                minimum=0.0,
+                maximum=1.0,
+                step=0.1,
+                value=1.0,
+                label="top_p",
+                render=False,
+            ),
+            gr.Slider(
+                minimum=1,
+                maximum=20,
+                step=1,
+                value=20,
+                label="top_k",
+                render=False,
+            ),
+            gr.Slider(
+                minimum=0.0,
+                maximum=2.0,
+                step=0.1,
+                value=1.2,
+                label="Repetition penalty",
+                render=False,
+            ),
+        ],
+        examples=[
+            ["Help me study vocabulary: write a sentence for me to fill in the blank, and I'll try to pick the correct option."],
+            ["What are 5 creative things I could do with my kids' art? I don't want to throw them away, but it's also so much clutter."],
+            ["Tell me a random fun fact about the Roman Empire."],
+            ["Show me a code snippet of a website's sticky header in CSS and JavaScript."],
+        ],
+        cache_examples=False,
+    )
+
+
+if __name__ == "__main__":
+    demo.launch()