more

app.py

@@ -1,25 +1,21 @@
 import gradio as gr
 from unsloth import FastLanguageModel
-from transformers import TextStreamer
 import torch
 
 # Load the model and tokenizer locally
 max_seq_length = 2048
-dtype = None
 model_name_or_path = "michailroussos/model_llama_8d"
 
 # Load model and tokenizer using unsloth
 model, tokenizer = FastLanguageModel.from_pretrained(
     model_name=model_name_or_path,
     max_seq_length=max_seq_length,
-    dtype=dtype,
     load_in_4bit=True,
 )
 FastLanguageModel.for_inference(model)  # Enable optimized inference
 
 # Define the response function
 def respond(message, history, system_message, max_tokens, temperature, top_p):
-    # Combine system message and conversation history
     messages = [{"role": "system", "content": system_message}]
     for user_msg, assistant_msg in history:
         if user_msg:
@@ -28,17 +24,15 @@ def respond(message, history, system_message, max_tokens, temperature, top_p):
             messages.append({"role": "assistant", "content": assistant_msg})
     messages.append({"role": "user", "content": message})
     
-    # Tokenize inputs
     inputs = tokenizer.apply_chat_template(
         messages,
         tokenize=True,
         add_generation_prompt=True,
         return_tensors="pt",
     ).to("cuda" if torch.cuda.is_available() else "cpu")
-
-    attention_mask = inputs.ne(tokenizer.pad_token_id).long()  # Explicitly set attention mask
-
-    # Generate response tokens
+    
+    attention_mask = inputs.ne(tokenizer.pad_token_id).long()
+    
     generated_tokens = model.generate(
         input_ids=inputs,
         attention_mask=attention_mask,
@@ -47,15 +41,9 @@ def respond(message, history, system_message, max_tokens, temperature, top_p):
         temperature=temperature,
         top_p=top_p,
     )
-
-    # Decode generated tokens
     response = tokenizer.decode(generated_tokens[0], skip_special_tokens=True)
-
-    # Yield response in the required Gradio format
     yield [{"role": "assistant", "content": response}]
 
-
-
 # Define the Gradio interface
 demo = gr.ChatInterface(
     respond,
@@ -65,7 +53,7 @@ demo = gr.ChatInterface(
         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)"),
     ],
-    type="messages"
+    type="messages",
 )
 
 if __name__ == "__main__":