Update app.py

app.py

@@ -5,9 +5,6 @@ sys_message = """
 This model can generate untruths, lies or inappropriate things. Only for testing and validation. 
 """
 
-# Load the model from Hugging Face Model Hub
-llm = AutoModelForCausalLM.from_pretrained("tevykuch/sl0th")
-
 # Configuration settings for model generation (example)
 generation_config = {
     "max_new_tokens": 2048,
@@ -18,15 +15,17 @@ generation_config = {
     "stop_token": '### Instruction:'
 }
 
-def stream(prompt, UL):
-    system_prompt = 'You are a helpful chatbot. You only answer in Khmer. User is based in Cambodia. Below is an instruction that describes a task, paired with an input that provides further context. Write a response that appropriately completes the request.'
-    E_INST = " "
-    system, user, assistant = "###Instruction", "###Input", "###Response"
-    prompt = f"{system}\n{system_prompt}{E_INST}\n{user}\n{prompt.strip()}{E_INST}\n{assistant}\n"
-
-    output = llm.generate(prompt, **generation_config)
-    return output
+tokenizer = AutoTokenizer.from_pretrained("tevykuch/sl0th")
+llm = AutoModelForCausalLM.from_pretrained("tevykuch/sl0th")
 
+def stream(prompt):
+    # Tokenize the prompt
+    inputs = tokenizer.encode(prompt, return_tensors="pt")
+    # Generate a response
+    output_ids = llm.generate(inputs, **generation_config)
+    # Decode the generated ids to a string
+    response = tokenizer.decode(output_ids[0], skip_special_tokens=True)
+    return response
 
 
 
@@ -44,16 +43,16 @@ chat_interface = gr.ChatInterface(
 )
 
 with gr.Blocks() as demo:
-    gr.HTML("<h1><center> sl0th <h1><center>")
+    gr.HTML("<h1><center> sl0th inference tester only (not final) <h1><center>")
     gr.HTML(
 			"<h4 style='text-align: center'>"
 				"<a href='https://huggingface.co/tevykuch/sl0th' target='_blank'>Model: Sl0th Mistral 7b 0.2</a> | "
 			"</h4>"
 		)
-    gr.HTML("<p><center>Finetune do <a href='https://huggingface.co/unsloth/mistral-7b-bnb-4bit' target='_blank'>Mistral 7b</a> com o dataset <a href='https://huggingface.co/datasets/dominguesm/alpaca-data-pt-br' target='_blank'>Alpaca-data-pt-br</a>.<p><center>")
+    gr.HTML("<p><center>Finetune here <a href='https://huggingface.co/unsloth/mistral-7b-bnb-4bit' target='_blank'>Mistral 7b</a> thanks dataset maker (my coworker) <a href='https://huggingface.co/datasets/metythorn/khmerllm-dataset-alpaca-52k-v1'>Alpaca-data-pt-br</a>.<p><center>")
     chat_interface.render()
     gr.Markdown(sys_message)
-    gr.DuplicateButton(value="Duplicar espaço para GPU", elem_id="duplicate-button")
+    gr.DuplicateButton(value="Duplicate the Magic", elem_id="duplicate-button")
     
 if __name__ == "__main__":
     demo.queue(max_size=10).launch()