import gradio as gr
import os
os.system('CMAKE_ARGS="-DLLAMA_OPENBLAS=on" FORCE_CMAKE=1 pip install llama-cpp-python')
import wget
from llama_cpp import Llama
import random
import os
import multiprocessing
from huggingface_hub import hf_hub_download #load from huggingfaces 

def get_num_cores():
    """Get the number of CPU cores."""
    return os.cpu_count()

def get_num_threads():
    """Get the number of threads available to the current process."""
    return multiprocessing.cpu_count()

if __name__ == "__main__":
    num_cores = get_num_cores()
    num_threads = get_num_threads()

    print(f"Number of CPU cores: {num_cores}")
    print(f"Number of threads available to the current process: {num_threads}")
#url = 'https://huggingface.co/TheBloke/Llama-2-7B-Chat-GGML/resolve/main/llama-2-7b-chat.ggmlv3.q2_K.bin'
#filename = wget.download(url)
model_path= hf_hub_download(repo_id="Aryanne/Bling-Sheared-Llama-1.3B-0.1-gguf", filename="q4_0-bling-sheared-llama-1.3b-0.1.gguf")

llm2 = Llama(model_path=model_path, seed=random.randint(1, 2**31), use_mlock=False, n_threads=4)
theme = gr.themes.Soft(
    primary_hue=gr.themes.Color("#ededed", "#fee2e2", "#fecaca", "#fca5a5", "#f87171", "#ef4444", "#dc2626", "#b91c1c", "#991b1b", "#7f1d1d", "#6c1e1e"),
    neutral_hue="red",
)
title = """<h1 align="center">Chat with awesome LLAMA 2 CHAT model!</h1><br>"""
with gr.Blocks(theme=theme) as demo:
    gr.HTML(title)
    gr.HTML("This model is awesome for its size! It is only 20th the size of Chatgpt but is still decent for chatting. However like all models, LLAMA-2-CHAT can hallucinate and provide incorrect information.")
    #chatbot = gr.Chatbot()
    #msg = gr.Textbox()
    #clear = gr.ClearButton([msg, chatbot])
    #instruction = gr.Textbox(label="Instruction", placeholder=)
    def bot(user_message):
        #token1 = llm.tokenize(b"### Instruction: ")
        #token2 = llm.tokenize(instruction.encode())
        #token3 = llm2.tokenize(b"USER: ")
        #tokens3 = llm2.tokenize(user_message.encode())
        #token4 = llm2.tokenize(b"\n\n### Response:")
        tokens = llm2.tokenize(user_message.encode())
        count = 0
        output = ""
        outputs = ""
        for token in llm2.generate(tokens, top_k=50, top_p=0.73, temp=0.72, repeat_penalty=1.1):
            text = llm2.detokenize([token])
            outputs += text.decode(errors='ignore')
            count += 1
            if count >= 500 or (token == llm2.token_eos()):
                break
            output += text.decode(errors='ignore')
            yield output
    gr.HTML("Thanks for checking out this app!")
    gr.Button("Answer").click(
        fn=bot, 
        inputs=gr.Textbox(),
        outputs=gr.Textbox(),
    )
demo.queue()
demo.launch(debug=True)