app.py

from transformers import AutoTokenizer
import gradio as gr


def tokenize(input_text):
    llama_tokens = len(llama_tokenizer(input_text, add_special_tokens=True)["input_ids"])
    llama3_tokens = len(llama3_tokenizer(input_text, add_special_tokens=True)["input_ids"])
    mistral_tokens = len(mistral_tokenizer(input_text, add_special_tokens=True)["input_ids"])
    gpt2_tokens = len(gpt2_tokenizer(input_text, add_special_tokens=True)["input_ids"])
    gpt_neox_tokens = len(gpt_neox_tokenizer(input_text, add_special_tokens=True)["input_ids"])
    falcon_tokens = len(falcon_tokenizer(input_text, add_special_tokens=True)["input_ids"])
    phi2_tokens = len(phi2_tokenizer(input_text, add_special_tokens=True)["input_ids"])
    phi3_tokens = len(phi3_tokenizer(input_text, add_special_tokens=True)["input_ids"])
    t5_tokens = len(t5_tokenizer(input_text, add_special_tokens=True)["input_ids"])
    gemma_tokens = len(gemma_tokenizer(input_text, add_special_tokens=True)["input_ids"])
    command_r_tokens = len(command_r_tokenizer(input_text, add_special_tokens=True)["input_ids"])
    qwen_tokens = len(qwen_tokenizer(input_text, add_special_tokens=True)["input_ids"])
    codeqwen_tokens = len(codeqwen_tokenizer(input_text, add_special_tokens=True)["input_ids"])
    rwkv4_tokens = len(rwkv4_tokenizer(input_text, add_special_tokens=True)["input_ids"])
    rwkv5_tokens = len(rwkv5_tokenizer(input_text, add_special_tokens=True)["input_ids"])
    deepseekv2_tokens = len(deepseekv2_tokenizer(input_text, add_special_tokens=True)["input_ids"])
    internlm_tokens = len(internlm_tokenizer(input_text, add_special_tokens=True)["input_ids"])
    internlm2_tokens = len(internlm2_tokenizer(input_text, add_special_tokens=True)["input_ids"])

    results = {
        "LLaMa-1/LLaMa-2": llama_tokens,
        "LLaMa-3": llama3_tokens,
        "Mistral": mistral_tokens,
        "GPT-2/GPT-J": gpt2_tokens,
        "GPT-NeoX": gpt_neox_tokens,
        "Falcon": falcon_tokens,
        "Phi-1/Phi-2": phi2_tokens,
        "Phi-3": phi3_tokens,
        "T5": t5_tokens,
        "Gemma": gemma_tokens,
        "Command-R": command_r_tokens,
        "Qwen/Qwen1.5": qwen_tokens,
        "CodeQwen": codeqwen_tokens,
        "RWKV-v4": rwkv4_tokens,
        "RWKV-v5/RWKV-v6": rwkv5_tokens,
        "DeepSeek-V2": deepseekv2_tokens,
        "InternLM": internlm_tokens,
        "InternLM2": internlm2_tokens
    }

    # Sort the results in descending order based on token length
    sorted_results = sorted(results.items(), key=lambda x: x[1], reverse=True)

    return "\n".join([f"{model}: {tokens}" for model, tokens in sorted_results])


if __name__ == "__main__":
    llama_tokenizer = AutoTokenizer.from_pretrained("TheBloke/Llama-2-7B-fp16")
    llama3_tokenizer = AutoTokenizer.from_pretrained("unsloth/llama-3-8b")
    mistral_tokenizer = AutoTokenizer.from_pretrained("mistral-community/Mistral-7B-v0.2")
    gpt2_tokenizer = AutoTokenizer.from_pretrained("gpt2")
    gpt_neox_tokenizer = AutoTokenizer.from_pretrained("EleutherAI/gpt-neox-20b")
    falcon_tokenizer = AutoTokenizer.from_pretrained("tiiuae/falcon-7b")
    phi2_tokenizer = AutoTokenizer.from_pretrained("microsoft/phi-2")
    phi3_tokenizer = AutoTokenizer.from_pretrained("microsoft/Phi-3-mini-4k-instruct")
    t5_tokenizer = AutoTokenizer.from_pretrained("google/flan-t5-xxl")
    gemma_tokenizer = AutoTokenizer.from_pretrained("alpindale/gemma-2b")
    command_r_tokenizer = AutoTokenizer.from_pretrained("PJMixers/CohereForAI_c4ai-command-r-plus-tokenizer", use_fast=False)
    qwen_tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen1.5-7B")
    codeqwen_tokenizer = AutoTokenizer.from_pretrained("Qwen/CodeQwen1.5-7B")
    rwkv4_tokenizer = AutoTokenizer.from_pretrained("RWKV/rwkv-4-14b-pile", trust_remote_code=True)
    rwkv5_tokenizer = AutoTokenizer.from_pretrained("RWKV/v5-EagleX-v2-7B-HF", trust_remote_code=True)
    deepseekv2_tokenizer = AutoTokenizer.from_pretrained("deepseek-ai/DeepSeek-V2", trust_remote_code=True)
    internlm_tokenizer = AutoTokenizer.from_pretrained("internlm/internlm-20b", trust_remote_code=True)
    internlm2_tokenizer = AutoTokenizer.from_pretrained("internlm/internlm2-20b", trust_remote_code=True)

    iface = gr.Interface(fn=tokenize, inputs=gr.Textbox(label="Input Text", lines=19), outputs="text")
    iface.launch()