[Rules] Dont post large blobs of text / code. Dont post install issues, use our google groups for them. Questions of the format: "I ran this code", "I got this error" are better off posted to google groups or github issues as they will be indexed by google search for the future.
Hello, I have a question regarding cudnn.benchmark. I don't find much information about it, only that "uses the inbuilt cudnn auto-tuner to find the fastest convolution algorithms". Also that if the batches are the same size then use it, otherwise it will make it slower. I want to know if this is helpful when I am running the code in inference time just once. It unfortunately reduces my speed like twice, so I'm wondering how and when should I use it?? And if there is any source to read more about it. Thanks
Hello, does torch7 supports nvidia 2080 Ti or nvidia binary driver-version 410.73?
Why are you killing Torch7 the best ml language ? I remember a time when things looked so promising.
@kenkit I hear you.. but pyTorch is actually really good.. if you haven't already, I'd check it out.. the biggest advantage from my point of view is the extensive libraries that are available in Python that just isn't there in Lua..
is possible to create a captcha solver? I have very difficult
@bnsh I like lua and anything that's C like, I hate python it's why I don't use Tensorflow which is much more popular. But Torch is still better than the rest since lua is simple and fun.
I hear you... Well... Minus the "hate python" part. It was nice having lua so close to C, and being able to just directly go into C whenever lua couldn't do what I needed... But, It's also nice to be able to use (for instance, numpy, or scipy or sklearn) for all the algorithms that they have, that neither Lua nor C had available to it...
Is anyone here at NIPS?
@bnsh I guess I 'll have to move where with the tide this time :(
Lua has always been a niche language and will stay like that. It is actually a miracle that Torch even exists knowing how few libraries with such broad scope Lua has in general.
I'm not a coder/developer, at all, but i find all this very interesting, especially for the art work processing on videos... I was wondering if Torch was, in fact, just the pilaster of what is Prisma app and Artisto app... Prisma Lab sales its SDK for 3000$ a month... I will never be able to develop myself a tool to process videos on my machine (linux), or even install "artistic-videos" (well i didn't try), I was wondering, how long and with how much efforts it's gone take for a talented team of developers to make a solid piece of software able to do what prisma and artisto do, on PC ? Or may be it's a wrong path, now that phones have the Neural Network processor unit like Kirin 970 to support AI, compare to the pc world... Thanks for reading... https://youtu.be/izEuz1K34Ro
Sorry, I should have said, the code exist, some solid pieces of software exist (artistic-videos, fast-artistic-videos), just missing GUI for instance..
@jsenellart I forgot to say thank you
I think it is important to keep it up and running
how i should install torch on win10?
hi all. May i know what is the license type for torch?
Python is fabulous and easily extended with C. CPython and Cython can speed anything up
Is there anybody here?
Nobody here but us chickens..
How to find out the name of the contributors of Torch?