Update app.py

app.py

@@ -1,14 +1,82 @@
-import spaces
+import os
+from threading import Thread
+from typing import Iterator
+
 import gradio as gr
-from huggingface_hub import InferenceClient
+import spaces
+import torch
+from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer
+
+
+MAX_INPUT_TOKEN_LENGTH = int(os.getenv("MAX_INPUT_TOKEN_LENGTH", "4096"))
+
+
+DESCRIPTION = """\
+# CataLlama-v0.1 Chat
+This Space demonstrates model [CataLlama-v0.1-Instruct-DPO](https://huggingface.co/catallama/CataLlama-v0.1-Instruct-DPO).
 
+CataLlama is a fine-tune on Llama-3 8B to enhance it's proficiency on the Catalan Language.
 """
-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
+
+if not torch.cuda.is_available():
+    DESCRIPTION += "\n<p>Running on CPU 🥶 This demo does not work on CPU.</p>"
+
+
+if torch.cuda.is_available():
+    model_id = "catallama/CataLlama-v0.1-Instruct-DPO"
+    model = AutoModelForCausalLM.from_pretrained(model_id, device_map="auto")
+    tokenizer = AutoTokenizer.from_pretrained(model_id)
+
+
+@spaces.GPU
+def generate(
+    message: str,
+    chat_history: list[tuple[str, str]],
+    system_prompt: 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 = []
+    if system_prompt:
+        conversation.append({"role": "system", "content": system_prompt})
+    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, 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=10.0, 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,
+        top_p=top_p,
+        top_k=top_k,
+        temperature=temperature,
+        num_beams=1,
+        repetition_penalty=repetition_penalty,
+    )
+    t = Thread(target=model.generate, kwargs=generate_kwargs)
+    t.start()
+
+    outputs = []
+    for text in streamer:
+        outputs.append(text)
+        yield "".join(outputs)
+
+
 """
 client = InferenceClient("catallama/CataLlama-v0.1-Instruct-DPO")
 
-
-@spaces.GPU(duration=120)
 def respond(
     message,
     history: list[tuple[str, str]],
@@ -40,12 +108,11 @@ def respond(
 
         response += token
         yield response
-
-"""
-For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
 """
-demo = gr.ChatInterface(
-    respond,
+
+
+chat_interface = gr.ChatInterface(
+    fn=generate,
     additional_inputs=[
         gr.Textbox(value="Ets un chatbot amigable.", label="System message"),
         gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
@@ -60,6 +127,10 @@ demo = gr.ChatInterface(
     ],
 )
 
+with gr.Blocks() as demo:
+    gr.Markdown(DESCRIPTION)
+    chat_interface.render()
+
 
 if __name__ == "__main__":
     demo.launch()