simple question about nllb-inference

[daje0601]

Hello, I saw your youtube and github code. So Thank you!

Thanks to this, I was able to run it in Colab. When I ran it, I was able to observe that it showed really good performance.

I want to customize this bash code to python. bash is so hard.. Can you tell me how to fix this code in python? If that's difficult, can you explain how this code works?

please merry on me

[pluiez]

Hi，the video is posted by @amrrs.

The script starts from translate.sh and run preprocess/normalize_punctuation.sh to normalize punctuation according to given langauge, then run spm_encode to encode the normalized text to sentence pieces. Finally, it runs fairseq's built-in command fairseq-interactve to receive inputs from standard input to generate translations.

By the way, Bash script is a nice tool to chain multiple programs as a pipeline, you might want to learn it if not yet.

[daje0601]

First of all, I really thank you for your response. I agree that Bash script is great. I will learn bash script later.

What I'm thinking about it that bash script always reload the model I want to just language translation. I don't catch how can I fix it? I'm still thinking..

[pluiez]

First of all, I really thank you for your response. I agree that Bash script is great. I will learn bash script later.

�What I'm thinking about it that bash script always reload the model I want to just language translation. I don't catch how can I fix it? I'm still thinking..

Do you mean you want to load a model and then change the langauges later for translation?

[daje0601]

First of all, I really thank you for your response. I agree that Bash script is great. I will learn bash script later. What I'm thinking about it that bash script always reload the model I want to just language translation. I don't catch how can I fix it? I'm still thinking..

Do you mean you want to load a model and then change the langauges later for translation?

Yes! Exactly!

[daje0601]

Actually, I am a junior who has only been learning Python for a year. However, we are growing by challenging myself as much as possible.

[pluiez]

First of all, I really thank you for your response. I agree that Bash script is great. I will learn bash script later. What I'm thinking about it that bash script always reload the model I want to just language translation. I don't catch how can I fix it? I'm still thinking..

Do you mean you want to load a model and then change the langauges later for translation?

Yes! Exactly!

You can check this https://stackoverflow.com/questions/66525973/running-fairseq-in-memory-and-pre-load-language-models

[daje0601]

Thank you!