i managed to get a deep learning algo running, and i am currently training it. i use torch, and its pretty hacked together.
i had to write my own matrix vector multiplication, that is very slow.
adding numpy would not only greatly simplify my programming, but also speed up my training and allow for bigger neural networks because of the time penalty.
You can just upload numpy with your algo.
I think that there is no reason to favor certain packages in certain programming languages by giving them an advantage on the server side.
sorry, i am a bit of a python noob, how can i upload numpy with my algo? and i am afraid i dont get what you want to say in the 2nd part of your response.
thanks for the fast reply
If the team decided to provide a selection of packages on their servers, I believe that it would give an unfair advantage to algos that make use of them or make it a requirement to use that package because you would lose otherwise.
Implementing an algo without it is a challenge on its own, which is why I cannot help you there.
1 Like
ok i understand. i considered numpy quite common and unspecific but i don’t know about the vast majority of its capability so i assume you are right.