znlp / bigtranslate Goto Github PK

Hi,

Thanks for contribution to LLM for MT. It seems that the paper/repo only shows the prompt for evaluation rather than translation. Is it possible to provide the prompts for translation for easier reproduction?

Cheers

Hi,

I have a question about the paper. In your paper, you stated "LLaMA that covers only 20 languages" in the abstract. May I ask where you obtained this number? And what languages are these 20?

Have been using and I modified the max_new tokens to be 2048 but the text never gets outputted fully. What is the maximum tokens that can be generated? I seem to be stuck on like 1-2 sentences max.

Hello, wanna ask 2 questions:

1. the way using in this repo is using prefix instruct like user have to send 把这个翻译为英语：xxx, why not just try system prompt like chatgpt, this can be used in multi-turn translation, for instance: 请扮演一个中英互译的翻译官，用户输入英文你翻译为对应的中文，xxx

2. Is there any place can download the translated data?

Hi,

It's a great job, thanks for your contribution to LLM. I've try to use your model before&after the update [June 06,2023], and I wonder how to use mutli-GPUs when inference?

In the new "translate.sh" , there are "export CUDA_VISIBLE_DEVICES=" . I set the value '3,4' , but it still runs on the single GPU 3.

Can the training code and data be released? Would like to see what is happening with numbers. It seems to lack the ability of translating sentences with numbers over 2020. Can you release the data?

我使用的prompt：
prompt1 = '''请将以下英语句子翻译成汉语：{}'''

prompt1 = '''请将以下句子翻译成汉语：{}'''

但模型没有输出，难受，不知道是哪里错了

Please release a 7B model, the 13B model on huggingface is not necessary and is an overkill for many simple use cases