www.digitalmars.com         C & C++   DMDScript  

digitalmars.D.announce - Tensorflow wrapper for D

reply 9il <ilyayaroshenko gmail.com> writes:
by Shigeki Karita

https://github.com/ShigekiKarita/tfd
May 31
next sibling parent Greatsam4sure <greatsam4sure yahoo.com> writes:
On Sunday, 31 May 2020 at 14:43:46 UTC, 9il wrote:
 by Shigeki Karita

 https://github.com/ShigekiKarita/tfd
Thanks for this great work. I will give it a try later. Tensorflow im D is whatwhile
May 31
prev sibling next sibling parent Greatsam4sure <greatsam4sure yahoo.com> writes:
On Sunday, 31 May 2020 at 14:43:46 UTC, 9il wrote:
 by Shigeki Karita

 https://github.com/ShigekiKarita/tfd
Thanks for this great work. I will give it a try later. Tensorflow in D is worthwhile
May 31
prev sibling next sibling parent jmh530 <john.michael.hall gmail.com> writes:
On Sunday, 31 May 2020 at 14:43:46 UTC, 9il wrote:
 by Shigeki Karita

 https://github.com/ShigekiKarita/tfd
Cool.
May 31
prev sibling next sibling parent Shigeki Karita <shigekikarita gmail.com> writes:
On Sunday, 31 May 2020 at 14:43:46 UTC, 9il wrote:
 by Shigeki Karita

 https://github.com/ShigekiKarita/tfd
Thanks for sharing my work. Currently, I am working on code generation for all the operations in https://github.com/tensorflow/tensorflow/blob/master/tensorflow/core/ops/ops.pbtxt
May 31
prev sibling parent reply tastyminerals <tastyminerals gmail.com> writes:
On Sunday, 31 May 2020 at 14:43:46 UTC, 9il wrote:
 by Shigeki Karita

 https://github.com/ShigekiKarita/tfd
Nice to see changes in DL domain for D. I wonder what happened to https://github.com/ShigekiKarita/grain, looks dormant now. Also, looks like tfd needs a lot of work too with autograd is not being implemented. Which TF version does tfd target? TF2 doesn't even have official Java API support which basically means that nobody will port their prod models to TF2 for the time being. I know that creating something akin to PyTorch/TF for D is a titanic amount of work. I wonder if generating PyTorch / TF bindings for D would be easier?
Jun 01
parent tastyminerals <tastyminerals gmail.com> writes:
On Monday, 1 June 2020 at 20:04:01 UTC, tastyminerals wrote:
 On Sunday, 31 May 2020 at 14:43:46 UTC, 9il wrote:
 by Shigeki Karita

 https://github.com/ShigekiKarita/tfd
Nice to see changes in DL domain for D. I wonder what happened to https://github.com/ShigekiKarita/grain, looks dormant now. Also, looks like tfd needs a lot of work too with autograd is not being implemented. Which TF version does tfd target? TF2 doesn't even have official Java API support which basically means that nobody will port their prod models to TF2 for the time being. I know that creating something akin to PyTorch/TF for D is a titanic amount of work. I wonder if generating PyTorch / TF bindings for D would be easier?
"nobody will port their prod models to TF2 for the time being." I meant those who use their TF models in Java/Scala code bases like we do.
Jun 01