Development Status :: 3 - Alpha

LLM Translator on Hugging-face models

LLMtranslator translates and generates text in multiple languages using LLMs(Large Language Models) on hugging-face models.

Introducing hf-transllm: Unlock the Power of Multilingual Exploration

Discover the hf-transllm package, a seamless integration of Hugging Face's inference module and translation APIs. Overcome limitations in retraining and evaluating large language models in different languages. Explore diverse results effortlessly, leveraging translation API services. Emphasizing versatility over efficiency, hf-transllm enables you to delve into the outcomes of Hugging Face's models in various languages.

Installation

pip install transllm

pip install git+https://github.com/dsdanielpark/hf-transllm.git

Quick Start

Simple Usage

from transllm import LLMtranslator

open_llama3b_kor = LLMtranslator('openlm-research/open_llama_3b', target_lang='ko', translator='google') # Korean

trnaslated_answer = open_llama3b_kor.generate("What's the weather in Seoul, Korea?")
print(trnaslated_answer)

If you wish to use CLI:

git clone https://github.com/dsdanielpark/hf-transllm
cd hf-transllm
pip install -r requirements.txt

python main.py --hfmodel <openlm-research/open_llama_3b> --lang <ko> --translator <google>

There can be issues with various dependencies such as Hugging Face's Transformers, SentencePiece, Torch, and CUDA. Please set up the appropriate environment.

Important

All Hugging Face models compatible. Check model card for template, modify as needed.

Translation API Integration Guide

Google Translator

Official use of Google Translate is chargeable. Use translator="google_official" and provide google_api_key. Unofficial testing should use translator="google". Refer to the official documentation for more details.

Official Google Translation API Supported Languages: Google Cloud Languages
Unofficial Google Translator (Non-profit/Testing) Supported Languages: Deep Translator Constants

Usage Example

from transllm import LLMtranslator

model_path = 'openlm-research/open_llama_3b'
# Alternative Models: 'openlm-research/open_llama_7b', 'openlm-research/open_llama_13b'

# For official Google Cloud Translation
# open_llama3b_kor = LLMtranslator(model_path, target_lang='ko', translator='google_official', google_api_key='YOUR_API_KEY')

# For unofficial testing
open_llama3b_kor = LLMtranslator(model_path, target_lang='ko', translator='google')

prompt = "Translate this text"
translated_answer = open_llama3b_kor.generate(prompt)
print(translated_answer)

DeepL

Supported Languages of DeepL: DeepL Languages

Pre-prompt Translation Example

from transllm import LLMtranslator

model_path = 'openlm-research/open_llama_3b'

# Choose your Translation Service API
open_llama3b_kor = LLMtranslator(model_path, target_lang='ES', translator='deepl', deepl_api='YOUR_DEEPL_API')
# Alternative setups for OpenAI and Bard

prompt = "Translate this text"
translated_response = open_llama3b_kor.generate(prompt)
print(translated_response)

OpenAI, Anthropic, Gemini Translation Services

In progress

Customized Inference

Customizing the inference process for unique prompt template or inference is possible. For advanced customization, add a translation module before and after the Hugging Face inference code.

Custom Inference Example

import torch
from trnasllm import LLMtranslator

class MyLLMtranslator(LLMtranslator):
    def __init__(self, model_path, target_lang="ko", translator="google", **kwargs):
        super().__init__(model_path=model_path, target_lang=target_lang, translator=translator, **kwargs)

    def inference(self, prompt: str) -> str:
        # Custom logic here
        return custom_logic(prompt)

FAQs

You can find most help on the FAQ and Issue pages.

Contributors

For detailed guidance on contributions, please refer to the contribution guide. We appreciate your interest in contributing and look forward to your valuable input.

Thank you for supporting our project.

License ©️

MIT license, 2024. We hereby strongly disclaim any explicit or implicit legal liability related to our works. Users are required to use this package responsibly and at their own risk. This project is a personal initiative and is not affiliated with or endorsed by Google, DeepL, Oepn AI and Anthropic.

Contacts

Reference

[1] https://huggingface.co/docs/api-inference/index

Copyright (c) 2023 MinWoo Park, South Korea

License

dsdanielpark/hf-transllm

About

Topics

Resources

License

Uh oh!

Stars

Watchers

Forks

Releases

Packages 0

Uh oh!

Contributors 2

Uh oh!

Languages

Packages