from ctransformers import AutoModelForCausalLM
import gradio as gr


llms = {
    "tinyllama":{"name": "TheBloke/TinyLlama-1.1B-1T-OpenOrca-GGUF", "file":"tinyllama-1.1b-1t-openorca.Q4_K_M.gguf", "suffix":"<|im_end|><|im_start|>assistant", "prefix":"<|im_start|>system You are a helpful assistant  <|im_end|><|im_start|>user"},
    "orca2":{"name": "TheBloke/Orca-2-7B-GGUF", "file":"orca-2-7b.Q4_K_M.gguf", "suffix":"<|im_end|><|im_start|>assistant", "prefix":"<|im_start|>system You are a helpful assistant<|im_end|><|im_start|>user "},
    "zephyr":{"name": "TheBloke/zephyr-7B-beta-GGUF", "file":"zephyr-7b-beta.Q4_K_M.gguf", "suffix":"</s><|assistant|>", "prefix":"<|system|>You are a helpful assistant</s><|user|> "},
    "mixtral":{"name": "TheBloke/Mistral-7B-Instruct-v0.1-GGUF", "file":"mistral-7b-instruct-v0.1.Q4_K_M.gguf", "suffix":"[/INST]", "prefix":"<s>[INST] "},
    "llama2":{"name": "TheBloke/Llama-2-7B-Chat-GGUF", "file":"llama-2-7b-chat.Q4_K_M.gguf", "suffix":"[/INST]", "prefix":"[INST] <<SYS>> You are a helpful assistant <</SYS>>"},
    "solar":{"name": "TheBloke/SOLAR-10.7B-Instruct-v1.0-GGUF", "file":"solar-10.7b-instruct-v1.0.Q4_K_M.gguf", "suffix":"\n### Assistant:\n", "prefix":"### User:\n"},
    #"open-llama": {"name": "TheBloke/open-llama-3b-v2-wizard-evol-instuct-v2-196k-GGUF", "file":"open-llama-3b-v2-wizard-evol-instuct-v2-196k.Q4_K_M.gguf", "suffix":"\n\n### RESPONSE", "prefix":"### HUMAN:\n"}
}

for k in llms.keys():
    AutoModelForCausalLM.from_pretrained(llms[k]['name'], model_file=llms[k]['file'])

import gradio as gr

def predict(prompt, llm_name):
    prefix=llms[llm_name]['prefix']
    suffix=llms[llm_name]['suffix']
    user="""
    {prompt}"""
    
    llm = AutoModelForCausalLM.from_pretrained(llms[llm_name]['name'], model_file=llms[llm_name]['file'])

    prompt = f"{prefix}{user.replace('{prompt}', prompt)}{suffix}"
    return llm(prompt)

# Create the Gradio interface
interface = gr.Interface(
    fn=predict,
    inputs=[gr.Textbox(label="Prompt", lines=20), gr.Dropdown(choices=list(llms), label="Select an LLM", value="tinyllama")],
    outputs="text"
)

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