Update space

app.py

@@ -1,10 +1,16 @@
 import gradio as gr
-from huggingface_hub import InferenceClient
+from transformers import AutoModelForCausalLM, AutoTokenizer
+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("Hack337/WavGPT-1.0")
+device = "cuda" if torch.cuda.is_available() else "cpu"
+
+model_path = "Hack337/WavGPT-1.0" # Replace with the actual model path
+model = AutoModelForCausalLM.from_pretrained(
+    model_path,
+    torch_dtype="auto",
+    device_map="auto"
+)
+tokenizer = AutoTokenizer.from_pretrained(model_path)
 
 
 def respond(
@@ -25,23 +31,28 @@ def respond(
 
     messages.append({"role": "user", "content": message})
 
-    response = ""
-
-    for message in client.chat_completion(
+    text = tokenizer.apply_chat_template(
         messages,
-        max_tokens=max_tokens,
-        stream=True,
+        tokenize=False,
+        add_generation_prompt=True
+    )
+    model_inputs = tokenizer([text], return_tensors="pt").to(device)
+
+    generated_ids = model.generate(
+        model_inputs.input_ids,
+        max_new_tokens=max_tokens,
+        pad_token_id=tokenizer.eos_token_id,
         temperature=temperature,
-        top_p=top_p,
-    ):
-        token = message.choices[0].delta.content
+        top_p=top_p
+    )
+    generated_ids = [
+        output_ids[len(input_ids):] for input_ids, output_ids in zip(model_inputs.input_ids, generated_ids)
+    ]
+    response = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
+
+    return response
 
-        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,
     additional_inputs=[

requirements.txt

@@ -1,2 +1,5 @@
 huggingface_hub==0.22.2
-minijinja
+minijinja
+torch
+transformers
+gradio