Update app.py

app.py

@@ -1,42 +1,106 @@
-# app.py
-
 import gradio as gr
-from transformers import AutoTokenizer, AutoModelForCausalLM
-import torch
-import numpy as np
-
-# Load the model and tokenizer
-model_name = "HuggingFaceH4/zephyr-7b-beta"  # Replace this with the actual model path
-tokenizer = AutoTokenizer.from_pretrained(model_name)
-model = AutoModelForCausalLM.from_pretrained(model_name)
-
-def get_category_scores(text):
-    # Tokenize input text
-    inputs = tokenizer(text, return_tensors="pt")
-    
-    # Generate model outputs
-    with torch.no_grad():
-        outputs = model(**inputs)
-    
-    # Process model output to extract labels
-    # For demonstration, we'll mock the extraction with random values
-    # Replace this part with actual logic based on model output
-    
-    # Example random values for demonstration (these should be extracted from model)
-    scores = np.random.randint(0, 10, size=13)  # Replace with actual logic
-    
-    # Convert scores to 13-digit code
-    result_code = ''.join(map(str, scores))
-    return result_code
-
-# Define the Gradio interface
-iface = gr.Interface(
-    fn=get_category_scores,
-    inputs=gr.Textbox(lines=2, placeholder="Enter text here..."),
-    outputs="text",
-    title="Text Labeling",
-    description="Label text data based on emotion temperature and categories. Returns a 13-digit code."
+from huggingface_hub import InferenceClient
+
+"""
+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,
+):
+    # Prepare the messages, starting with the system message
+    messages = [{"role": "system", "content": system_message}]
+
+    # Add the conversation history to the messages
+    for user_message, assistant_response in history:
+        if user_message:
+            messages.append({"role": "user", "content": user_message})
+        if assistant_response:
+            messages.append({"role": "assistant", "content": assistant_response})
+
+    # Add the current user message
+    messages.append({"role": "user", "content": message})
+
+    response = ""
+
+    # Stream the response from the model
+    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
+
+
+"""
+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 tasked with labeling text data based on both emotion temperature and text type categories. The final output must be a 13-character code that consists of the following structure:
+
+1. Emotion Temperature Code (2 characters):
+   - If the emotion is purely Cold: Use CC
+   - If the emotion is purely Warm: Use WW
+   - If the emotion is purely Hot: Use HH
+   - If the emotion is a mix, use one of the following:
+     - Cold and Warm: Use CW
+     - Warm and Hot: Use WH
+     - Cold and Hot: Use CH
+
+2. Text Type Codes (next 9 digits):
+   Assign a digit for each of the following categories based on the presence in the text. Use 0 for categories not applicable:
+   1: Toxic
+   2: Appreciation
+   3: Constructive Criticism
+   4: Genuine Questions
+   5: Advice/Suggestions
+   6: Requests
+   7: Spam
+   8: Off-Topic
+   9: Engagement Boosters
+
+3. Special Categories (last 2 digits):
+   If the text is Neutral/General: Set the 10th digit to 1; otherwise, set it to 0.
+   If the text contains Hate: Set the last digit (11th) to 1; otherwise, set it to 0.
+
+Example:
+For the text "I love your videos but still something is missing":
+   - Emotion: Cold and Warm (CW)
+   - Types Detected: 2 (Appreciation), 3 (Constructive Criticism), 5 (Advice/Suggestions)
+   - Special Categories: Neutral/General (set the 10th digit to 1), no Hate
+
+The output would be: CW02305000010
+
+Output Format:
+Always return a 13-character code following this structure.""",
+            label="Instruction",
+            lines=10,
+        ),
+        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)",
+        ),
+    ],
 )
 
-# Launch the interface
-iface.launch()
+if __name__ == "__main__":
+    demo.launch()