text2dataset

Easily turn large English text datasets into Japanese text datasets using open LLMs.

A tool for converting a datasets.Dataset by translating the data in the "txt" column using Open LLM like gemma2 with vLLM, and adding a new "txt_ja" column (translated text in Japanese). This tool is inspired by img2dataset.

Features

• Save the intermediate results in shards:
  • By setting the number_sample_per_shard parameter, the dataset can be saved in shards as specified by the number of samples per shard.
• Resume from checkpoint:
  • By setting the resume_from_checkpoint parameter, the translation can be resumed from where it left off.
• Logging with wandb:
  • By setting the use_wandb parameter, the metrics such as examples_per_sec and count can be logged to wandb.
• Push to Hugging Face Hub:
  • By setting the push_to_hub parameter, the translated dataset can be pushed to the Hugging Face Hub.

Usage

$ python src/text2dataset/main.py \
    --model_id "google/gemma-2-9b-it" \
    --batch_size 16384 \
    --input_format parquet \
    --input_path "/path/to/input" \
    --source_column "caption" \
    --target_column "caption_ja" \
    --push_to_hub False \
    --push_to_hub_path "/path/to/hub" \
    --output_dir "/path/to/output" \
    --output_format parquet \
    --gpu_id 0 \
    --number_sample_per_shard 10000 \
    --use_wandb True

Example

You can use Translator class to translate texts into Japanese.

from datasets import load_dataset
from text2dataset.translator import Translator

ds = load_dataset("Abirate/english_quotes", split="train")
ds = ds.select(range(10))
print(ds.column_names)
# ['quote', 'author', 'tags']
print("\n".join(ds["quote"][:5]))
# “Be yourself; everyone else is already taken.”
# “I'm selfish, impatient and a little insecure. I make mistakes, I am out of control and at times hard to handle. But if you can't handle me at my worst, then you sure as hell don't deserve me at my best.”
# “Two things are infinite: the universe and human stupidity; and I'm not sure about the universe.”
# “So many books, so little time.”
# “A room without books is like a body without a soul.”
translator = Translator(model_id="google/gemma-2-9b-it", tensor_parallel_size=1, pipeline_parallel_size=1)
translated = translator.translate(ds["quote"])
ds = ds.add_column("quote_ja", translated)
print(ds.column_names)
# ['quote', 'author', 'tags', 'quote_ja']
print("\n".join(ds["quote_ja"][:5]))
#
# 自分のことは、自己中心的で、衝動的で、少し不安定。失敗することもあるし、制御不能な時もあるし、扱いにくい時もある。でも、私が最悪な時をあなたが処理できないなら、最高の私をあなたが望む資格はない。
# **宇宙と人間の愚かさ、どちらが無限大か分からない。**
# 本がたくさん、時間が足りない。
# 書籍のない部屋は、魂のない体と同じ。

Areas for Improvement

• Data Paarallel Inference:
  • Currently, only one model is used for inference. This can be improved by using DataParallel. If you know how to do this with vLLM, please let me know or Pull Request.

References

• https://github.com/vllm-project/vllm
• https://github.com/rom1504/img2dataset