import gradio as gr
from gpt4all import GPT4All
from huggingface_hub import hf_hub_download

title = "DiarizationLM GGUF inference on CPU"

description = """
DiarizationLM GGUF inference on CPU
"""

model_path = "models"
# model_name = "model-unsloth.Q4_K_M.gguf"
model_name = "model-unsloth.BF16.gguf"
hf_hub_download(repo_id="google/DiarizationLM-13b-Fisher-v1", filename=model_name, local_dir=model_path, local_dir_use_symlinks=False)

print("Start the model init process")
model = model = GPT4All(model_name, model_path, allow_download = False, device="cpu")
print("Finish the model init process")

model.config["promptTemplate"] = "{0} --> "
model.config["systemPrompt"] = ""
model._is_chat_session_activated = False

max_new_tokens = 2048

print("Finish the model config process")

def generater(message, history, temperature, top_p, top_k):
    prompt = model.config["promptTemplate"].format(message)
    outputs = []    
    for token in model.generate(prompt=prompt, temp=temperature, top_k = top_k, top_p = top_p, max_tokens = max_new_tokens, streaming=True):
        outputs.append(token)
        yield "".join(outputs)


def vote(data: gr.LikeData):
    if data.liked:
        return
    else:
        return

print("Create chatbot")
chatbot = gr.Chatbot()
print("Created chatbot")

print("Add additional_inputs")
additional_inputs=[
    gr.Slider(
        label="temperature",
        value=0.0,
        minimum=0.0,
        maximum=2.0,
        step=0.05,
        interactive=True,
        info="Higher values like 0.8 will make the output more random, while lower values like 0.2 will make it more focused and deterministic.",
    ),
    gr.Slider(
        label="top_p",
        value=1.0,
        minimum=0.0,
        maximum=1.0,
        step=0.01,
        interactive=True,
        info="0.1 means only the tokens comprising the top 10% probability mass are considered. Suggest set to 1 and use temperature. 1 means 100% and will disable it",
    ),
    gr.Slider(
        label="top_k",
        value=50,
        minimum=0,
        maximum=1000,
        step=1,
        interactive=True,
        info="limits candidate tokens to a fixed number after sorting by probability. Setting it higher than the vocabulary size deactivates this limit.",
    )
]
print("Added additional_inputs")

iface = gr.ChatInterface(
    fn = generater,
    title=title,
    description = description,
    chatbot=chatbot,
    additional_inputs=additional_inputs,
    examples=[
        ["<speaker:1> Hello, how are you doing <speaker:2> today? I am doing well."],
   ]
)

print("Added iface")

with gr.Blocks() as demo:
    chatbot.like(vote, None, None)
    iface.render()

print("Rendered iface")

if __name__ == "__main__":
    demo.queue(max_size=3).launch()