python is good for prototyping and understanding.
\nit is easy to debug and step through
but it is slow.
\ni think today language model is good enough to make model for translation of python to other language like C/c++.
\nmaybe kaggle can organize competition to translate python code to faster executable.
\nthe leaderboard metric can be speed or correctness of results.
for example:
\nhttps://ai.facebook.com/blog/deep-learning-to-translate-between-programming-languages/
python is good for prototyping and understanding.
\nit is easy to debug and step through
but it is slow.
\ni think today language model is good enough to make model for translation of python to other language like C/c++.
\nmaybe kaggle can organize competition to translate python code to faster executable.
\nthe leaderboard metric can be speed or correctness of results.
for example:
\nhttps://ai.facebook.com/blog/deep-learning-to-translate-between-programming-languages/
Good Idea. Is translation cost justifiable? is translated code well readable by humans? because while translating from one programming language to another simple logic sometimes becomes complex.
", "rawMarkdown": "Good Idea. Is translation cost justifiable? is translated code well readable by humans? because while translating from one programming language to another simple logic sometimes becomes complex.", "votes": null }, { "id": "1807900", "postDate": "06/01/2022 12:34:34", "content": "The python in pytorch/tensorflow is just an interface, the actual code for language and vision models is C/C++ and Cuda-C. Also PyTorch has a C++ interface so there is no really need in converting.
", "rawMarkdown": "The python in pytorch/tensorflow is just an interface, the actual code for language and vision models is C/C++ and Cuda-C. Also PyTorch has a C++ interface so there is no really need in converting.", "votes": null }, { "id": "1815101", "postDate": "06/08/2022 17:10:45", "content": "There is always a tradeoff between performance and efficiency. Everything depends on user needs as well as implementation overhead.
", "rawMarkdown": "There is always a tradeoff between performance and efficiency. Everything depends on user needs as well as implementation overhead.", "votes": null }, { "id": "1820940", "postDate": "06/15/2022 05:50:40", "content": "Yup, today most of the popular python frameworks use C++/cuda implementations behind the scenes and also devlopment in python is realtively easy.
", "rawMarkdown": "Yup, today most of the popular python frameworks use C++/cuda implementations behind the scenes and also devlopment in python is realtively easy.", "votes": null }, { "id": "1822192", "postDate": "06/16/2022 06:13:46", "content": "If text data is insufficient in real life, data is augmented through back-translation. Thank you for the good idea!
", "rawMarkdown": "If text data is insufficient in real life, data is augmented through back-translation. Thank you for the good idea!", "votes": null } ], "comments": [ { "id": 1804726, "author_name": "muhammedtausif", "author_url": "", "post_date": "05/29/2022 11:38:24", "content": "Good Idea. Is translation cost justifiable? is translated code well readable by humans? because while translating from one programming language to another simple logic sometimes becomes complex.
", "votes": null, "replies": [] }, { "id": 1807900, "author_name": "skyyy93", "author_url": "", "post_date": "06/01/2022 12:34:34", "content": "The python in pytorch/tensorflow is just an interface, the actual code for language and vision models is C/C++ and Cuda-C. Also PyTorch has a C++ interface so there is no really need in converting.
", "votes": null, "replies": [ { "id": 1820940, "author_name": "klokeshsai", "author_url": "", "post_date": "06/15/2022 05:50:40", "content": "Yup, today most of the popular python frameworks use C++/cuda implementations behind the scenes and also devlopment in python is realtively easy.
", "votes": null, "replies": [] } ] }, { "id": 1815101, "author_name": "ashrafr", "author_url": "", "post_date": "06/08/2022 17:10:45", "content": "There is always a tradeoff between performance and efficiency. Everything depends on user needs as well as implementation overhead.
", "votes": null, "replies": [] }, { "id": 1822192, "author_name": "heuingttiang", "author_url": "", "post_date": "06/16/2022 06:13:46", "content": "If text data is insufficient in real life, data is augmented through back-translation. Thank you for the good idea!
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "1804589": "python is good for prototyping and understanding.\nit is easy to debug and step through\n\nbut it is slow.\n\ni think today language model is good enough to make model for translation of python to other language like C/c++.\n\nmaybe kaggle can organize competition to translate python code to faster executable.\nthe leaderboard metric can be speed or correctness of results.\n\nfor example:\nhttps://ai.facebook.com/blog/deep-learning-to-translate-between-programming-languages/", "1804726": "Good Idea. Is translation cost justifiable? is translated code well readable by humans? because while translating from one programming language to another simple logic sometimes becomes complex.", "1807900": "The python in pytorch/tensorflow is just an interface, the actual code for language and vision models is C/C++ and Cuda-C. Also PyTorch has a C++ interface so there is no really need in converting.", "1815101": "There is always a tradeoff between performance and efficiency. Everything depends on user needs as well as implementation overhead.", "1820940": "Yup, today most of the popular python frameworks use C++/cuda implementations behind the scenes and also devlopment in python is realtively easy.", "1822192": "If text data is insufficient in real life, data is augmented through back-translation. Thank you for the good idea!" }, "source": "meta" }