add gpu option #674
add gpu option #674
Conversation
|
LGTM. @wannaphong this is one of my Builders. Should we tell him to fix the PEP8 before merge? |
Yes, I think like you. |
|
I don't sure that It can working with GPU by use CUDA_VISIBLE_DEVICES to set available GPU. If it can use by use CUDA_VISIBLE_DEVICES to set available GPU, I think maybe we needs pull request to update docs not add gpu parameter. |
pythainlp/translate/core.py
Outdated
| """ | ||
| :param str src_lang: source language | ||
| :param str target_lang: target language | ||
| :param str target_lang: target |
There was a problem hiding this comment.
Can you add use_gpu parameter to the docs?
pythainlp/translate/en_th.py
Outdated
| @@ -51,7 +51,7 @@ class EnThTranslator: | |||
|
|
|||
| Website: https://airesearch.in.th/releases/machine-translation-models/ | |||
| """ | |||
| def __init__(self): | |||
| def __init__(self, use_gpu: bool): | |||
There was a problem hiding this comment.
Can you add use_gpu parameter to the docs?
pythainlp/translate/en_th.py
Outdated
| @@ -51,7 +51,7 @@ class EnThTranslator: | |||
|
|
|||
| Website: https://airesearch.in.th/releases/machine-translation-models/ | |||
| """ | |||
| def __init__(self): | |||
| def __init__(self, use_gpu: bool): | |||
There was a problem hiding this comment.
I think it should has default is False.
|
I've resolved all your comments. And should I add docs to the init of each class? |
pythainlp/translate/core.py
Outdated
| @@ -7,6 +7,7 @@ class Translate: | |||
|
|
|||
| :param str src_lang: source language | |||
| :param str target_lang: target language | |||
| :param bool use_gpu: load model to gpu | |||
There was a problem hiding this comment.
I think you should add default. example: (Default is False)
There was a problem hiding this comment.
Okay, added to all of them
See [3.1 Milestone](https://github.com/PyThaiNLP/pythainlp/milestone/16). ## What is new? ### Deprecation and other API changes #687 Remove deprecated function - pythainlp.word_vector; doesnt_match, get_model, most_similar_cosmul, sentence_vectorizer, similarity. use WordVector class instead - pythainlp.util.delete_tone. use pythainlp.util.remove_tonemark instead - Remove pythainlp.util.time_time. use pythainlp.util.time_to_thaiword instead - pythainlp.tokenize.syllable_tokenize. use pythainlp.tokenize.subword_tokenize instead ### Name Entity Tagging - #665 Add Thai-NNER `pythainlp.tag.NNER` - #658 Add LST20NER onnx model. It is LST20NER model to onnx model from fine-turning by [WangchanBERTa model](https://huggingface.co/airesearch/wangchanberta-base-att-spm-uncased). ### Transliteration - #659 Add ISO 11940 transliteration - #660 Add Thai W2P v0.2 - #686 Add wunsen - #694 Wunsen Mandarin and Japanese update ### PyThaiNLP Corpus downloader - #656 Add support zip/tar.gz to download corpus ### Text normalization - #673 Add a normalising rule for Lakkhangyao ๅ ### Translate - #674 add gpu option ### Text summarize - #679 Add mt5 cpe kmutt thai sentence sum ### Util - #682 Add live-dead syllable classification - #684 Add live dead syllable classify - #690 Add tone detector ### Other - #689 map NG tag to PART - #691 Remove TinyDB as a dependency - #692 Fix notifications that newer versions of corpora are available
What does this changes
add gpu option on translate module