Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Repo info
Activity
  • Dec 01 16:05

    drahnr on bernhard-chores-2022

    (compare)

  • Dec 01 16:05

    drahnr on master

    bump rust-bindgen to 0.60.1, bu… (compare)

  • Dec 01 16:05
    drahnr closed #174
  • Dec 01 16:04

    drahnr on arch-refactor

    Coaster convolution API cleanup… (compare)

  • Dec 01 16:04
    drahnr closed #178
  • Nov 19 19:47
    hweom edited #178
  • Nov 19 19:46
    hweom synchronize #178
  • Nov 05 22:25
    hweom edited #178
  • Nov 05 22:24
    hweom opened #178
  • Oct 25 14:50

    drahnr on arch-refactor

    Add trainer subsystem with SGD … (compare)

  • Oct 25 14:50
    drahnr closed #177
  • Oct 24 15:34
    drahnr synchronize #177
  • Oct 24 15:34
    drahnr synchronize #177
  • Oct 23 18:27
    hweom synchronize #177
  • Oct 23 18:26
    hweom opened #177
  • Oct 23 07:40

    drahnr on arch-refactor

    Add a few more layers to the ne… (compare)

  • Oct 23 07:40
    drahnr closed #176
  • Oct 22 20:55
    hweom synchronize #176
  • Oct 10 02:17
    hweom edited #176
  • Oct 10 02:17
    hweom edited #176
colepoirier
@colepoirier:matrix.org
[m]
BatchNormalization2d is the first layer that it uses
So once I follow the process you outlined for adding a new layer to juice and go through the PR and review process I plan on moving through the remainder of the layers that that model uses from pytorch and adding them from rcudnn-sys in juice
Obviously I will do each as a separate PR and follow proper git commit practice
Would you also like me to raise a corresponding issue for each layer before submitting the PR? Or is just the PR sufficient?
Bernhard Schuster
@drahnr
Yeah, adding individual issues would be much appreciated! Thank you
I'll try to review your PRs promptly
colepoirier
@colepoirier:matrix.org
[m]
Ok cool :)
colepoirier
@colepoirier:matrix.org
[m]
I ended up having quite a hectic week with IRL appointments and commitments. So I’ll finally be setting up my desktop with the 1080ti and new power supply tomorrow (Saturday) morning and should be on my way on the code shortly thereafter if all the cuDNN installation etc goes well
Bernhard Schuster
@drahnr
No rush at all!
colepoirier
@colepoirier:matrix.org
[m]
:)
colepoirier
@colepoirier:matrix.org
[m]
Everything is installed and the examples compile! So I’m all set to start now... well I’ll start working on juice tomorrow actually :)
The spearow.io link on the github repo errors with 502 bad gateway. Are you already aware of this?
Bernhard Schuster
@drahnr
No, I wasn't, I'll take care of it! Thanks
colepoirier
@colepoirier:matrix.org
[m]
Ok great! I’m interested to see what’s on the website!
Bernhard Schuster
@drahnr
I guess the default should be the book, or use a subdomain, I have to figure if github can handle this - before I had custom infra which I started to scale down a bit.
colepoirier
@colepoirier:matrix.org
[m]
Cool! It works 😎
What is spearow?
Bernhard Schuster
@drahnr
It's just a fictional name.
It's a mix of spearou (a Belgian comic) and sparrow (bird)
colepoirier
@colepoirier:matrix.org
[m]
Cool! Is it meant to be the name of an organization for machine learning?
Bernhard Schuster
@drahnr
The org
colepoirier
@colepoirier:matrix.org
[m]
Cool!
I like the name a lot!
colepoirier
@colepoirier:matrix.org
[m]
I started the task of implementing BatchNormalization in rcudnn last night and a thought occurred to me. Might it be easier to auto generate all of the ParamaterNotSupported error strings by parsing them per function from the cuDNN API reference pdf?
Whoops I meant “CUDNN_STATUS_BAD_PARAM”
Bernhard Schuster
@drahnr
Technically yes, but I think the effort required to impl such a tool would be very high compared to doing 10 to 100 manual error strings
colepoirier
@colepoirier:matrix.org
[m]
Fair enough that makes sense :)
colepoirier
@colepoirier:matrix.org
[m]
Hi @drahnr, I’m about halfway through the process you laid out for implementing BatchNormalization in juice but have become stuck with regard to implementing the ‘descriptor’ step. It seems like the non-LRN/non-divisive normalization cuDNN functions are quite different in their usage API than functions like LRN and Activation functions. Despite spending a solid half an hour reading the cuDNN API reference trying to decipher the correct way to implement this in juice I’m quite confused and stuck. Can we arrange to have a synchronous chat in the next few days so I can hopefully figure this out and proceed with implementing this set of cuDNN functions in juice? Cheers :)
Bernhard Schuster
@drahnr
Sure, how about Friday or Saturday?
Preferably in the evening hours
2 replies
Could create a draft PR upfront with your current changes?
Bernhard Schuster
@drahnr
@colepoirier:matrix.org ^^^
colepoirier
@colepoirier:matrix.org
[m]
Hey sorry
Sure will do when I get home today!
colepoirier
@colepoirier:matrix.org
[m]
I just need to know your timezone :)
I’m PST/PDT
Bernhard Schuster
@drahnr
I am CEST
colepoirier
@colepoirier:matrix.org
[m]
Ok great
Are you available at some point this afternoon or evening? It’s about 14:30 your time right now correct?
Also sorry I didn’t create the PR yet, I’ll do that today
Bernhard Schuster
@drahnr
I'd like to first have a peek into the PR
I'll have some time tomorrow, 20:00 CEST
colepoirier
@colepoirier:matrix.org
[m]
That makes sense, I just submitted the draft PR. Great! I saw your private message and have replied there
Aurelien Montmejat
@aurelien-m
Hey everyone, quick question : is greenglass the way to go for loading images to neural nets ?
Bernhard Schuster
@drahnr
Well, you could.
But it's meant as a preprocessing layer.
If it fits the bill, use it, if not, custom image decoding code might be faster
Aurelien Montmejat
@aurelien-m
Oh okay
Bernhard Schuster
@drahnr
Make sure to use the git version of juice