Update app.py

app.py

@@ -1,63 +1,79 @@
+from threading import Thread
+
 import gradio as gr
-from huggingface_hub import InferenceClient
+import spaces
+import torch
+from PIL import Image
+from transformers import AutoModelForVision2Seq, AutoProcessor, AutoTokenizer, TextIteratorStreamer
 
+
+TITLE = "<h1><center>Chat with PaliGemma-3B-Chat-v0.1</center></h1>"
+
+DESCRIPTION = "<h3><center>Visit <a href='https://huggingface.co/hiyouga/PaliGemma-3B-Chat-v0.1' target='_blank'>our model page</a> for details.</center></h3>"
+
+CSS = """
+.duplicate-button {
+  margin: auto !important;
+  color: white !important;
+  background: black !important;
+  border-radius: 100vh !important;
+}
 """
-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")
 
 
-def respond(
-    message,
-    history: list[tuple[str, str]],
-    system_message,
-    max_tokens,
-    temperature,
-    top_p,
-):
-    messages = [{"role": "system", "content": system_message}]
+model_id = "hiyouga/PaliGemma-3B-Chat-v0.1"
+tokenizer = AutoTokenizer.from_pretrained(model_id)
+processor = AutoProcessor.from_pretrained(model_id)
+model = AutoModelForVision2Seq.from_pretrained(model_id, torch_dtype="auto", device_map="auto")
+
 
-    for val in history:
-        if val[0]:
-            messages.append({"role": "user", "content": val[0]})
-        if val[1]:
-            messages.append({"role": "assistant", "content": val[1]})
+@spaces.GPU
+def stream_chat(message: Dict[str, str], history: list):
+    print(message)
+    conversation = []
+    for prompt, answer in history:
+        conversation.extend([{"role": "user", "content": prompt}, {"role": "assistant", "content": answer}])
 
-    messages.append({"role": "user", "content": message})
+    conversation.append({"role": "user", "content": message})
 
-    response = ""
+    input_ids = tokenizer.apply_chat_template(conversation, add_generation_prompt=True, return_tensors="pt").to(
+        model.device
+    )
+    streamer = TextIteratorStreamer(tokenizer, timeout=10.0, skip_prompt=True, skip_special_tokens=True)
 
-    for message in client.chat_completion(
-        messages,
-        max_tokens=max_tokens,
-        stream=True,
+    generate_kwargs = dict(
+        input_ids=input_ids,
+        streamer=streamer,
+        max_new_tokens=max_new_tokens,
         temperature=temperature,
-        top_p=top_p,
-    ):
-        token = message.choices[0].delta.content
+        do_sample=True,
+    )
+    if temperature == 0:
+        generate_kwargs["do_sample"] = False
 
-        response += token
-        yield response
+    t = Thread(target=model.generate, kwargs=generate_kwargs)
+    t.start()
 
-"""
-For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
-"""
-demo = gr.ChatInterface(
-    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"),
-        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
-        gr.Slider(
-            minimum=0.1,
-            maximum=1.0,
-            value=0.95,
-            step=0.05,
-            label="Top-p (nucleus sampling)",
-        ),
-    ],
-)
+    output = ""
+    for new_token in streamer:
+        output += new_token
+        yield output
+
+
+chatbot = gr.Chatbot(height=450)
+
+with gr.Blocks(css=CSS) as demo:
+    gr.HTML(TITLE)
+    gr.HTML(DESCRIPTION)
+    gr.DuplicateButton(value="Duplicate Space for private use", elem_classes="duplicate-button")
+    gr.ChatInterface(
+        fn=stream_chat,
+        multimodal=True,
+        chatbot=chatbot,
+        fill_height=True,
+        cache_examples=False,
+    )
 
 
 if __name__ == "__main__":
-    demo.launch()
+    demo.launch()