qwen

app.py

@@ -3,11 +3,10 @@ from huggingface_hub import InferenceClient
 import os
 
 # Securely retrieve the token from your Space secrets
-# Ensure you have a secret named HF_TOKEN in your Settings
 HF_TOKEN = os.getenv("HF_TOKEN")
 
-# Initialize the inference client with the specified model
-client = InferenceClient("HuggingFaceH4/zephyr-7b-beta", token=HF_TOKEN)
+# Initializing with a model that has high serverless availability
+client = InferenceClient("Qwen/Qwen2.5-7B-Instruct", token=HF_TOKEN)
 
 class StateController:
     def __init__(self):
@@ -44,9 +43,9 @@ def generate_response(message, history):
         output = "Diagnostic sequence initiated.\n\n"
         output += f"{controller.initialize_grid()}\n\n"
         output += "Rendering 121-point array:\n"
-        output += f"`{controller.render_grid()}`\n\n"
+        output += f"{controller.render_grid()}\n\n"
         output += "Executing state resolution:\n"
-        output += f"`{controller.resolve_grid()}`"
+        output += f"{controller.resolve_grid()}"
         return output
 
     system_instruction = (
@@ -55,35 +54,40 @@ def generate_response(message, history):
         "Provide direct, technical, and accurate responses."
     )
     
-    # Formatting for Gradio 6.5+ message history
-    formatted_messages = [{"role": "system", "content": system_instruction}]
-    for turn in history:
-        formatted_messages.append({"role": "user", "content": turn[0]})
-        formatted_messages.append({"role": "assistant", "content": turn[1]})
-    formatted_messages.append({"role": "user", "content": message})
+    # Building the prompt for the text generation API
+    prompt = f"<|im_start|>system\n{system_instruction}<|im_end|>\n"
+    for user_msg, assistant_msg in history:
+        prompt += f"<|im_start|>user\n{user_msg}<|im_end|>\n<|im_start|>assistant\n{assistant_msg}<|im_end|>\n"
+    prompt += f"<|im_start|>user\n{message}<|im_end|>\n<|im_start|>assistant\n"
 
     try:
-        response_text = ""
-        # Direct call for response generation
-        completion = client.chat_completion(
-            formatted_messages,
-            max_tokens=1024,
-            stream=False # Set to False for maximum stability during testing
+        # Using text_generation for direct inference API access
+        response = client.text_generation(
+            prompt,
+            max_new_tokens=1024,
+            stop_sequences=["<|im_end|>", "<|endoftext|>"],
+            temperature=0.1 # Low temperature for high precision
         )
-        return completion.choices[0].message.content
+        return response.strip()
     except Exception as error:
-        return f"System Error: {str(error)}. Ensure HF_TOKEN is correctly set in Secrets."
+        return f"System Error: {str(error)}. If the error persists, ensure your HF_TOKEN has Inference permissions."
 
-# Professional UI implementation
-with gr.Blocks(theme=gr.themes.Soft(primary_hue="blue")) as demo:
+custom_css = """
+body, .gradio-container { background-color: #0b0f19 !important; }
+footer {display: none !important}
+.message.user { background-color: #1e293b !important; border: 1px solid #3b82f6 !important; }
+.message.bot { background-color: #0f172a !important; color: #60a5fa !important; }
+"""
+
+with gr.Blocks(theme=gr.themes.Soft(primary_hue="blue"), css=custom_css) as demo:
     gr.Markdown("# Advanced Logic Interface")
     gr.ChatInterface(
         fn=generate_response,
         description="Inference layer utilizing state-hold logic.",
         examples=[
             "Run grid diagnostic",
-            "Explain network latency without using the word delay.",
-            "Calculate allocation for 120 units across 3 nodes."
+            "Calculate the integer distribution for 120 units across 3 nodes.",
+            "Explain network latency using technical terminology."
         ]
     )