Update app.py

app.py

@@ -1,49 +1,58 @@
 import gradio as gr
-from huggingface_hub import InferenceClient
+from transformers import LlamaTokenizer, LlamaForCausalLM
+import torch
 
-"""
-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("HuggingFaceH4/zephyr-7b-beta")
+model_repo_id = "Bllossom/llama-3-Korean-Bllossom-70B"
+
+# 토크나이저 로드
+tokenizer = LlamaTokenizer.from_pretrained(
+    model_repo_id,
+    use_auth_token='your_hf_access_token'  # 필요한 경우 액세스 토큰 추가
+)
+
+# 모델 로드
+model = LlamaForCausalLM.from_pretrained(
+    model_repo_id,
+    torch_dtype=torch.float16,  # 또는 torch.bfloat16
+    device_map="auto",          # 가능한 경우 GPU에 자동 할당
+    use_auth_token='your_hf_access_token'  # 필요한 경우 액세스 토큰 추가
+)
 
 def respond(
     message,
-    history: list[tuple[str, str]],
+    history,
     system_message,
     max_tokens,
     temperature,
     top_p,
 ):
-    messages = [{"role": "system", "content": system_message}]
+    # 프롬프트 생성
+    prompt = system_message + "\n"
+    for user_msg, bot_msg in history:
+        prompt += f"User: {user_msg}\nAssistant: {bot_msg}\n"
+    prompt += f"User: {message}\nAssistant:"
 
-    for val in history:
-        if val[0]:
-            messages.append({"role": "user", "content": val[0]})
-        if val[1]:
-            messages.append({"role": "assistant", "content": val[1]})
+    inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
 
-    messages.append({"role": "user", "content": message})
-
-    response = ""
-
-    for message in client.chat_completion(
-        messages,
-        max_tokens=max_tokens,
-        stream=True,
+    outputs = model.generate(
+        **inputs,
+        max_new_tokens=max_tokens,
         temperature=temperature,
         top_p=top_p,
-    ):
-        token = message.choices[0].delta.content
+        do_sample=True,
+        eos_token_id=tokenizer.eos_token_id,
+        pad_token_id=tokenizer.eos_token_id,
+    )
 
-        response += token
-        yield response
+    response = tokenizer.decode(outputs[0], skip_special_tokens=True)
+    response = response[len(prompt):].strip()
 
+    history.append((message, response))
+
+    return history
 
-"""
-For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
-"""
 demo = gr.ChatInterface(
-    respond,
+    fn=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"),
@@ -58,6 +67,5 @@ demo = gr.ChatInterface(
     ],
 )
 
-
 if __name__ == "__main__":
     demo.launch()