Update app.py

app.py

@@ -1,6 +1,5 @@
 import transformers
 import torch
-import gradio as gr
 
 model_id = "yodayo-ai/nephra_v1.0"
 
@@ -9,81 +8,63 @@ pipeline = transformers.pipeline(
     model=model_id,
     model_kwargs={"torch_dtype": torch.bfloat16},
     device_map="auto",
-    offload_folder="offload",  # Add this line to specify the offload folder
+    offload_folder="offload",  # Only provided here during model initialization
 )
 
-# Define characters
+# Define characters and traits
 characters = [
-    {
-        "name": "Alex",
-        "description": "Alex is a young and ambitious adventurer, full of energy and eager to discover new places. He is always ready to face any challenges and strives to explore the unknown.",
-        "traits": "brave, energetic, optimistic, determined"
-    },
-    {
-        "name": "Maya",
-        "description": "Maya is a wise and experienced sorceress, possessing deep knowledge of magic and ancient rituals. She is known for her calm demeanor, analytical mind, and ability to find solutions in difficult situations.",
-        "traits": "calm, thoughtful, intuitive, attentive"
-    },
-    {
-        "name": "Victor",
-        "description": "Victor is a former warrior who left behind his fighting days to seek inner peace and harmony. His life experience and sense of justice make him a reliable friend and mentor.",
-        "traits": "serious, judicious, fair, balanced"
-    }
+    {"name": "Alex", 
+     "description": "Alex is a young and ambitious adventurer, full of energy and a thirst for new discoveries. Always ready to face any challenge, he is driven by a desire to explore uncharted places.", 
+     "traits": "brave, energetic, optimistic, determined"},
+    
+    {"name": "Maya", 
+     "description": "Maya is a wise and experienced sorceress, with deep knowledge in magic and ancient rituals. She is known for her calm demeanor, analytical mind, and ability to find solutions in difficult situations.",
+     "traits": "calm, thoughtful, intuitive, attentive"},
+    
+    {"name": "Victor", 
+     "description": "Victor is a former warrior who gave up fighting for inner peace and harmony. His life experience and pursuit of justice make him a reliable friend and mentor.", 
+     "traits": "serious, thoughtful, fair, balanced"}
 ]
 
-# Function to generate response
-def generate_response(character_name, user_input, max_length=100, temperature=1.12, min_p=0.075, repetition_penalty=1.1):
-    # Find the character
+def generate_response(character_name, user_input, max_length=100, temperature=0.7, top_p=0.85, repetition_penalty=1.1):
+    # Find the character data
     character = next((c for c in characters if c["name"] == character_name), None)
     if not character:
         return "Character not found."
 
-    # Prepare the message based on the selected character's personality and description
-    messages = [
-        {"role": "system", "content": f"You are {character_name}, {character['description']}. Personality traits: {character['traits']}."},
-        {"role": "user", "content": user_input},
-    ]
-
-    # Prepare the prompt using the chat template from the pipeline's tokenizer
-    prompt = pipeline.tokenizer.apply_chat_template(
-        messages,
-        tokenize=False,
-        add_generation_prompt=True
-    )
+    # Create the prompt text
+    prompt_text = (f"You are {character_name}, {character['description']}. Traits: {character['traits']}. "
+                   f"In response to the question '{user_input}', respond {random.choice(['inspired', 'with doubt', 'joyfully', 'thoughtfully', 'skeptically'])}. Please complete the response.")
 
-    # Generate response using the pipeline
+    # Generate response
     outputs = pipeline(
-        prompt,
+        prompt_text,
         max_new_tokens=max_length,
-        eos_token_id=[
-            pipeline.tokenizer.convert_tokens_to_ids(""),
-            pipeline.tokenizer.eos_token_id,
-        ],
         do_sample=True,
         temperature=temperature,
-        min_p=min_p,
-        repetition_penalty=repetition_penalty,
+        top_p=top_p,
+        repetition_penalty=repetition_penalty
     )
-
-    # Extract and return the generated response
-    generated_text = outputs[0]["generated_text"][len(prompt):].strip()
-    return generated_text
+    
+    return outputs[0]['generated_text']
 
 # Gradio Interface
+import gradio as gr
+
 iface = gr.Interface(
     fn=generate_response,
     inputs=[
-        gr.Dropdown([c["name"] for c in characters], label="Choose a character"),
+        gr.Dropdown([c["name"] for c in characters], label="Choose Character"),
         gr.Textbox(lines=2, placeholder="Enter your text here..."),
         gr.Slider(20, 200, step=1, value=100, label="Max Length"),
-        gr.Slider(0.1, 1.0, step=0.1, value=1.12, label="Temperature"),
-        gr.Slider(0.01, 1.0, step=0.01, value=0.075, label="min-p"),
+        gr.Slider(0.1, 1.0, step=0.1, value=0.7, label="Temperature"),
+        gr.Slider(0.1, 1.0, step=0.05, value=0.85, label="Top-p"),
         gr.Slider(1.0, 2.0, step=0.1, value=1.1, label="Repetition Penalty")
     ],
     outputs="text",
-    title="Nephra v1 LLM Roleplaying Chatbot",
-    description="Enter a text prompt to generate a response using the Nephra v1 model, based on the selected character."
+    title="Roleplaying Model Demo",
+    description="Generate responses based on the chosen character's traits."
 )
 
 if __name__ == "__main__":
-    iface.launch(share=True, debug=True)
+    iface.launch()