import subprocess # Install flash attention, skipping CUDA build if necessary subprocess.run( "export CPATH=$CPATH:/usr/local/cuda/include", shell=True, ) subprocess.run( "export LIBRARY_PATH=$LIBRARY_PATH:/usr/local/cuda/lib64", shell=True, ) subprocess.run( "/home/user/.pyenv/shims/pip install flash-attn --no-build-isolation", env={"FLASH_ATTENTION_SKIP_CUDA_BUILD": "TRUE"}, shell=True, ) subprocess.run( "/home/user/.pyenv/shims/pip install pycuda==2023.1", env={"CPATH": "$CPATH:/usr/local/cuda/include", "LIBRARY_PATH": "$LIBRARY_PATH:/usr/local/cuda/lib64"}, shell=True, ) import gradio as gr import os import spaces from transformers import AutoModelForCausalLM from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer from threading import Thread # Set an environment variable HF_TOKEN = os.environ.get("HF_TOKEN", None) DESCRIPTION = '''

Meta Llama3 8B

This Space demonstrates the instruction-tuned model Meta Llama3 8b Chat. Meta Llama3 is the new open LLM and comes in two sizes: 8b and 70b. Feel free to play with it, or duplicate to run privately!

šŸ”Ž For more details about the Llama3 release and how to use the model with transformers, take a look at our blog post.

šŸ¦• Looking for an even more powerful model? Check out the Hugging Chat integration for Meta Llama 3 70b

''' LICENSE = """

--- Built with Meta Llama 3 """ PLACEHOLDER = """

Meta llama3

Ask me anything...

""" css = """ h1 { text-align: center; display: block; } #duplicate-button { margin: auto; color: white; background: #1565c0; border-radius: 100vh; } """ # Load the tokenizer and model model_name = "gradientai/Llama-3-8B-Instruct-262k" tokenizer = AutoTokenizer.from_pretrained(model_name) model = AutoModelForCausalLM.from_pretrained(model_name, torch_dtype="auto", device_map="auto") # to("cuda:0") from minference import MInference minference_patch = MInference("minference", model_name) model = minference_patch(model) terminators = [ tokenizer.eos_token_id, tokenizer.convert_tokens_to_ids("<|eot_id|>") ] @spaces.GPU(duration=120) def chat_llama3_8b(message: str, history: list, temperature: float, max_new_tokens: int ) -> str: """ Generate a streaming response using the llama3-8b model. Args: message (str): The input message. history (list): The conversation history used by ChatInterface. temperature (float): The temperature for generating the response. max_new_tokens (int): The maximum number of new tokens to generate. Returns: str: The generated response. """ # global model conversation = [] for user, assistant in history: conversation.extend([{"role": "user", "content": user}, {"role": "assistant", "content": assistant}]) conversation.append({"role": "user", "content": message}) input_ids = tokenizer.apply_chat_template(conversation, return_tensors="pt").to(model.device) print(model.device) # subprocess.run( # "pip install pycuda==2023.1", # shell=True, # ) # if "has_patch" not in model.__dict__: # from minference import MInference # minference_patch = MInference("minference", model_name) # model = minference_patch(model) streamer = TextIteratorStreamer(tokenizer, timeout=10.0, skip_prompt=True, skip_special_tokens=True) generate_kwargs = dict( input_ids= input_ids, streamer=streamer, max_new_tokens=max_new_tokens, do_sample=True, temperature=temperature, eos_token_id=terminators, ) # This will enforce greedy generation (do_sample=False) when the temperature is passed 0, avoiding the crash. if temperature == 0: generate_kwargs['do_sample'] = False t = Thread(target=model.generate, kwargs=generate_kwargs) t.start() outputs = [] for text in streamer: outputs.append(text) #print(outputs) yield "".join(outputs) # Gradio block chatbot=gr.Chatbot(height=450, placeholder=PLACEHOLDER, label='Gradio ChatInterface') with gr.Blocks(fill_height=True, css=css) as demo: gr.Markdown(DESCRIPTION) gr.DuplicateButton(value="Duplicate Space for private use", elem_id="duplicate-button") gr.ChatInterface( fn=chat_llama3_8b, chatbot=chatbot, fill_height=True, additional_inputs_accordion=gr.Accordion(label="āš™ļø Parameters", open=False, render=False), additional_inputs=[ gr.Slider(minimum=0, maximum=1, step=0.1, value=0.95, label="Temperature", render=False), gr.Slider(minimum=128, maximum=4096, step=1, value=512, label="Max new tokens", render=False ), ], examples=[ ['How to setup a human base on Mars? Give short answer.'], ['Explain theory of relativity to me like Iā€™m 8 years old.'], ['What is 9,000 * 9,000?'], ['Write a pun-filled happy birthday message to my friend Alex.'], ['Justify why a penguin might make a good king of the jungle.'] ], cache_examples=False, ) gr.Markdown(LICENSE) if __name__ == "__main__": demo.launch()