app.py

# ADD DISCLAIMERS
# AND LOGGING
# Q: why is this model so fucking slow? A: because i'm not made of money

import gradio as gr
import os
os.system("apt install libopenblas-dev")
os.system("make clean && LLAMA_OPENBLAS=1 make")
os.system('CMAKE_ARGS="-DLLAMA_BLAS=ON -DLLAMA_BLAS_VENDOR=OpenBLAS" pip install llama-cpp-python transformers')

from llama_cpp import Llama
from transformers import AutoTokenizer
model_id = "Elijahbodden/eliGPTv1.1"

# MODEL
model = Llama.from_pretrained(
    repo_id=model_id,
    filename="eliGPTv1.1-unsloth.Q5_K_M.gguf",
    verbose=True,
    n_threads = 2,
    n_threads_batch = 2,
    n_ctx=8192,
  )


# TOKENIZER AND TEMPLATE
tokenizer = AutoTokenizer.from_pretrained(model_id)

sys_prompt = """SUMMARY - ELIJAH:
Age: 16
Interests: space flight, cybernetics, consciousness, philosophy, psychonautism, biotech, AI
Likes: thinking and learning, building stuff, interesting conversations, red hot chili peppers and techno, humanism
Traits: incredibly intelligent, funny, interesting, caffeine fiend, very ambitious, militant atheist, self-taught/homeschooled, casual
Aspirations: creating transhumanist utopia, understanding the universe more, becoming smarter and better\n"""


your_name = "elijah"
custom_template = \
    f"{{% if {'sys_prompt' in locals()} %}}"\
      "{{ '<|im_start|>system\n" + sys_prompt + "<|im_end|>\n' }}"\
    "{% endif %}"\
    "{% for message in messages %}"\
        "{% if message['role'] == 'user' %}"\
            "{{'<|im_start|>user\n' + message['content'] + '<|im_end|>\n'}}"\
        "{% elif message['role'] == 'assistant' %}"\
            "{{'<|im_start|>" + your_name + "\n' + message['content'] + eos_token + '<|im_end|>\n' }}"\
        "{% else %}"\
            "{{ '<|im_start|>system\n' + message['content'] + '<|im_end|>\n' }}"\
        "{% endif %}"\
    "{% endfor %}"\
    "{{ '<|im_start|>" + your_name + "\n' }}"\

tokenizer.chat_template = custom_template

presets = {
    "emojis" : [{"role": "user", "content": "Wait I like you so much more when you use emojis more, keep doing it 😂\n"}, {"role": "assistant", "content": "Ummm, ok, looks like i'll be using more emojis 💀\n"}],
    "none" : []
}


def respond(
    message,
    history: list[tuple[str, str]],
    max_tokens,
    temperature,
    mirostat_tau,
    mirostat_eta,
):
    preset = "emojis"
    
    messages = presets[preset]
    for val in history:
        if val[0]:
            messages.append({"role": "user", "content": val[0]})
        if val[1]:
            messages.append({"role": "assistant", "content": val[1]})

    messages.append({"role": "user", "content": message})

    response = ""

    convo = tokenizer.apply_chat_template(messages, tokenize=False)
    print(convo)
    for message in model.create_completion(
        convo,
        temperature=0.75,
        stream=True,
        stop=["<|im_end|>"],
        mirostat_mode=1,
        mirostat_tau=mirostat_tau,
        mirostat_eta=mirostat_eta,
        max_tokens=128
    ):
        token = message["choices"][0]["text"]

        response += token
        yield response

"""
For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
"""
demo = gr.ChatInterface(
    respond,
    additional_inputs_accordion="The juicy stuff (settings)",
    css=".bubble-gap {gap: 6px !important}",
    # description="EliGPT v1.idon'tfuckingknow",
    title="EliGPT v1.idon'tfuckingknow",
    additional_inputs=[
        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
        gr.Slider(
            minimum=0.0,
            maximum=10.0,
            value=3.0,
            step=0.5,
            label="Mirostat tau",
        ),
        gr.Slider(
            minimum=0.0,
            maximum=1.0,
            value=0.1,
            step=0.01,
            label="Mirostat eta",
        ),
    ],
)


if __name__ == "__main__":
    demo.launch()