Update app.py

app.py

@@ -1,70 +1,61 @@
+import gradio as gr
 import transformers
 import torch
 
+# Model and pipeline setup
 model_id = "yodayo-ai/nephra_v1.0"
 
-pipeline = transformers.pipeline(
-    "text-generation",
-    model=model_id,
-    model_kwargs={"torch_dtype": torch.bfloat16},
+model = transformers.AutoModelForCausalLM.from_pretrained(
+    model_id,
+    torch_dtype=torch.bfloat16,
     device_map="auto",
-    offload_folder="offload",  # Only provided here during model initialization
+    offload_folder="offload"  # Ensure this folder is available or adjust the path
 )
 
-# Define characters and traits
-characters = [
-    {"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"}
-]
+tokenizer = transformers.AutoTokenizer.from_pretrained(model_id)
 
-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."
+pipeline = transformers.pipeline(
+    "text-generation",
+    model=model,
+    tokenizer=tokenizer,
+    device_map="auto",
+)
 
-    # 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.")
+# Function to generate a response
+def generate_response(user_input):
+    messages = [
+        {"role": "system", "content": "You are to play the role of a cheerful assistant."},
+        {"role": "user", "content": user_input},
+    ]
+
+    prompt = pipeline.tokenizer.apply_chat_template(
+        messages,
+        tokenize=False,
+        add_generation_prompt=True
+    )
 
-    # Generate response
     outputs = pipeline(
-        prompt_text,
-        max_new_tokens=max_length,
+        prompt,
+        max_new_tokens=512,
+        eos_token_id=[
+            pipeline.tokenizer.convert_tokens_to_ids(""),
+            pipeline.tokenizer.eos_token_id,
+        ],
         do_sample=True,
-        temperature=temperature,
-        top_p=top_p,
-        repetition_penalty=repetition_penalty
+        temperature=1.12,
+        min_p=0.075,
     )
     
-    return outputs[0]['generated_text']
+    return outputs[0]["generated_text"][len(prompt):]
 
 # Gradio Interface
-import gradio as gr
-
-iface = gr.Interface(
+interface = gr.Interface(
     fn=generate_response,
-    inputs=[
-        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=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")
-    ],
+    inputs="text",
     outputs="text",
-    title="Roleplaying Model Demo",
-    description="Generate responses based on the chosen character's traits."
+    title="Chat with Nephra",
+    description="Interact with the Nephra model, a roleplaying and instruction-based AI.",
 )
 
-if __name__ == "__main__":
-    iface.launch()
+# Launch the Gradio app
+interface.launch()