app.py

import gradio as gr
from tokenizers import Tokenizer

llama3_tokenizer = Tokenizer.from_file("tokenizer-llama3.json")
deepseek_coder_tokenizer = Tokenizer.from_file("tokenizer-deepseek-coder.json")

def get_tokenizer(model):
    tokenizer_mapping = {
        "meta-llama/Meta-Llama-3-8B-Instruct": llama3_tokenizer, 
        "deepseek-ai/deepseek-coder-7b-instruct-v1.5": deepseek_coder_tokenizer
    }
    if model not in tokenizer_mapping:
        raise Exception(f"Model {model} not supported.")

    return tokenizer_mapping[model]

def count_tokens(
    model,
    target_text,
):
    print("###############################")
    print(model)
    tokenizer = get_tokenizer(model)
    print(tokenizer)
    toks = tokenizer.encode(target_text)
    print(toks.ids)
    yield len(toks.ids)
   
def greet(name, intensity):
    return "Hello, " + name + "!" * int(intensity)

demo = gr.Interface(
    fn=count_tokens,
    inputs=[
        gr.Dropdown(
            [
                "meta-llama/Meta-Llama-3-8B-Instruct", 
                "deepseek-ai/deepseek-coder-7b-instruct-v1.5", 
            ], label="Model"
        ),
        gr.Textbox(
            label="Input",
            info="Text to count tokens for",
            lines=10,
        ),
    ],
    outputs=["text"],
)



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