Update app.py

app.py

@@ -1,72 +1,100 @@
+3.85 kB
+from transformers import pipeline, TextIteratorStreamer
+import torch
+from threading import Thread
 import gradio as gr
-from huggingface_hub import InferenceClient
+import spaces
+import re
 
-import os
-from huggingface_hub import InferenceClient
-
-# Read token from Space secrets
-hf_token = os.getenv("HF_TOKEN")
-
-client = InferenceClient("HuggingFaceH4/zephyr-7b-beta", token=hf_token)
-
-"""
-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}]
-
-    for val in history:
-        if val[0]:
-            messages.append({"role": "user", "content": val[0]})
-        if val[1]:
-            messages.append({"role": "assistant", "content": val[1]})
-
-    messages.append({"role": "user", "content": message})
-
-    response = ""
-
-    for message in client.chat_completion(
-        messages,
-        max_tokens=max_tokens,
-        stream=True,
-        temperature=temperature,
-        top_p=top_p,
-    ):
-        token = message.choices[0].delta.content
-
-        response += token
-        yield response
+model_id = "openai/gpt-oss-20b"
 
+pipe = pipeline(
+    "text-generation",
+    model=model_id,
+    torch_dtype="auto",
+    device_map="auto",
+)
+def format_conversation_history(chat_history):
+    messages = []
+    for item in chat_history:
+        role = item["role"]
+        content = item["content"]
+        if isinstance(content, list):
+            content = content[0]["text"] if content and "text" in content[0] else str(content)
+        messages.append({"role": role, "content": content})
+    return messages
+    
+@spaces.GPU()
+def generate_response(input_data, chat_history, max_new_tokens, system_prompt, temperature, top_p, top_k, repetition_penalty):
+    new_message = {"role": "user", "content": input_data}
+    system_message = [{"role": "system", "content": system_prompt}] if system_prompt else []
+    processed_history = format_conversation_history(chat_history)
+    messages = system_message + processed_history + [new_message]
+    streamer = TextIteratorStreamer(pipe.tokenizer, skip_prompt=True, skip_special_tokens=True)
+    generation_kwargs = {
+        "max_new_tokens": max_new_tokens,
+        "do_sample": True,
+        "temperature": temperature,
+        "top_p": top_p,
+        "top_k": top_k,
+        "repetition_penalty": repetition_penalty,
+        "streamer": streamer
+    }
+    thread = Thread(target=pipe, args=(messages,), kwargs=generation_kwargs)
+    thread.start()
+    # simple formatting without harmony because of no tool usage etc. and experienced hf space problems with harmony 
+    thinking = ""
+    final = ""
+    started_final = False
+    for chunk in streamer:
+        if not started_final:
+            if "assistantfinal" in chunk.lower():
+                split_parts = re.split(r'assistantfinal', chunk, maxsplit=1)
+                thinking += split_parts[0]
+                final += split_parts[1]
+                started_final = True
+            else:
+                thinking += chunk
+        else:
+            final += chunk
+        clean_thinking = re.sub(r'^analysis\s*', '', thinking).strip()
+        clean_final = final.strip()
+        formatted = f"<details open><summary>Click to view Thinking Process</summary>\n\n{clean_thinking}\n\n</details>\n\n{clean_final}"
+        yield formatted
 
-"""
-For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
-"""
 demo = gr.ChatInterface(
-    respond,
+    fn=generate_response,
     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)",
+        gr.Slider(label="Max new tokens", minimum=64, maximum=4096, step=1, value=2048),
+        gr.Textbox(
+            label="System Prompt",
+            value="You are a helpful assistant. Reasoning: medium",
+            lines=4,
+            placeholder="Change system prompt"
         ),
+        gr.Slider(label="Temperature", minimum=0.1, maximum=2.0, step=0.1, value=0.7),
+        gr.Slider(label="Top-p", minimum=0.05, maximum=1.0, step=0.05, value=0.9),
+        gr.Slider(label="Top-k", minimum=1, maximum=100, step=1, value=50),
+        gr.Slider(label="Repetition Penalty", minimum=1.0, maximum=2.0, step=0.05, value=1.0)
+    ],
+    examples=[
+        [{"text": "Explain Newton laws clearly and concisely"}],
+        [{"text": "Write a Python function to calculate the Fibonacci sequence"}],
+        [{"text": "What are the benefits of open weight AI models"}],
     ],
+    cache_examples=False,
+    type="messages",
+    description="""# gpt-oss-20b Demo
+Give it a couple of seconds to start. You can adjust reasoning level in the system prompt like "Reasoning: high." Click to view thinking process (default is on).""",
+    fill_height=True,
+    textbox=gr.Textbox(
+        label="Query Input",
+        placeholder="Type your prompt"
+    ),
+    stop_btn="Stop Generation",
+    multimodal=False,
+    theme=gr.themes.Soft()
 )
 
-
 if __name__ == "__main__":
-    demo.launch()
+    demo.launch(share=True)