by

Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Activity
  • 20:48

    Oceania2018 on master

    sigmoid_grad eager. (compare)

  • 20:47

    Oceania2018 on master

    Fully Connected Neural Network … (compare)

  • 15:40
    wensioness starred SciSharp/TensorFlow.NET
  • 13:50
    muyu7777 starred SciSharp/TensorFlow.NET
  • 13:50
  • 13:32

    Oceania2018 on master

    MNIST CNN (Eager) (compare)

  • 13:31

    Oceania2018 on master

    Change default stddev to 0.05f … (compare)

  • 12:59
    Oceania2018 commented #570
  • 12:59
    Oceania2018 labeled #570
  • 12:58
    Oceania2018 assigned #570
  • 12:58
    Oceania2018 assigned #570
  • 12:56
    Oceania2018 assigned #587
  • 12:55
    Oceania2018 labeled #587
  • 12:55
    Oceania2018 closed #587
  • 12:55
    Oceania2018 commented #587
  • 12:54
    Oceania2018 labeled #572
  • 12:54
    Oceania2018 assigned #572
  • 12:54
    Oceania2018 closed #573
  • 12:53
    Oceania2018 commented #573
  • 12:50
    Oceania2018 labeled #597
Gonzalo Casas
@gnz_twitter
hi folks, could you please hint me in the right direction? I would like to get NumSharp to work on ironpython. My first obvious try was simply to download the released DLL, do clr.AddReferenceToFileAndPath('NumSharp.Core.dll') and play with it. Importing the namespace works (import NumSharp), but trying any operation fails with Could not load file or assembly 'System.Memory, Version=4.0.1.1, Culture=neutral, PublicKeyToken=cc7b13ffcd2ddd51' or one of its dependencies. Is there any obvious step I am missing?
Gonzalo Casas
@gnz_twitter
ah.. nevermind, found it! I was really missing the correct System.Memory version. ;)
Marcin Borawski
@nlhnt
Hey guys, does NumSharp support complex numbers?
Meinrad Recheis
@henon

https://twitter.com/Haiping2015/status/1265730586938953730

that's awesome. better performance than Python is a very good USP for TFNET

Иван Сердюк
@oceanfish81_twitter
Jean-Sylvain Boige
@jsboige
Hi guys, I have been following SciSharp's stack development and want to thank you for your effort. Beforehand, when Cesar De Souza bailed out from Accord.Net and Keras Sharp, I was quite desperate for a follow up and (naively) placed Bounties for advancing the latter (https://www.bountysource.com/teams/keras-sharp/issues?tracker_ids=67797984). I just received a notification from Bountysource, offering to redeploy those bounties before the end of the month or they'll be cached out. Can you point me to issues in TF.Net or Keras sharp, where it would be useful to redeploy those bounties? Cheers
Jean-Sylvain Boige
@jsboige
meant Keras.Net sorry
Иван Сердюк
@oceanfish81_twitter

Hi guys, I have been following SciSharp's stack development and want to thank you for your effort. Beforehand, when Cesar De Souza bailed out from Accord.Net and Keras Sharp, I was quite desperate for a follow up and (naively) placed Bounties for advancing the latter (https://www.bountysource.com/teams/keras-sharp/issues?tracker_ids=67797984). I just received a notification from Bountysource, offering to redeploy those bounties before the end of the month or they'll be cached out. Can you point me to issues in TF.Net or Keras sharp, where it would be useful to redeploy those bounties? Cheers

It is interesting that you had something to do with Lucene.NET

Jean-Sylvain Boige
@jsboige
Well, this is mainly what I had to do with it: https://archive.codeplex.com/?p=ariciels, which ages a bit now.
Meinrad Recheis
@henon
@jsboige the maintainer of TF.Net is @Oceania2018 , the maintainer of Keras.NET is @deepakkumar1984. I hope they'll respond
Haiping
@Oceania2018
@jsboige Sorry, what's bounty? How can I help?
Nguyen Hoang Nguyen
@nguyenhn
Hi guys, I'm new with tensorflow.NET. I would like to use tensorflow.NET gpu version with tensorflow2.2 but the existing tfnet gpu nuget (https://www.nuget.org/packages/SciSharp.TensorFlow.Redist-Windows-GPU/) is just for tf 1.15.1 (Cuda 10) ofcourse it's out of date with tensorflowNET documentation. Any suggestion? Many thanks.
Jean-Sylvain Boige
@jsboige
Hi @Oceania2018, hi @deepakkumar1984 (I believe we already discussed about the state of SiaNet before you switched to Keras.Net, congrats for doing the right thing and not reimplementing the whole stack like you nearly went for). A bounty on boutysource is a cash prize that will go to whoever fixes the corresponding github issue. There is a significant amount to redeploy here, so it would be nice if you can point me to issues, that you know you'll eventually fix, and for which you'd be equally happy if a bounty hunter from bountysource fixes those issues first and get the prize.
Haiping
@Oceania2018
@nguyenhn Thanks for asking. tf.net 0.20 for tensorflow 2.0 is still in working progress, havn't provided GPU version yet.
Jean-Sylvain Boige
@jsboige
Hi, I'm sorry to insist with my request for bounties redeployment, but if I don't act quickly, those prizes will go to waste. Can someone with a good understanding of the stack direct me to significant github issues to put those bounties on?
Meinrad Recheis
@henon
@Oceania2018 is the man
chriss2401
@chriss2401
Any updates on the bounties ? Would be a pity to see them go to waste
Jean-Sylvain Boige
@jsboige
Hi @chriss2401, @Oceania2018 has pointed me to that issue: SciSharp/SciSharp-Stack-Examples#17, and I'm contemplating adding SciSharp/TensorFlow.NET#350 and SciSharp/TensorFlow.NET#265 if that works for you. I'm currently trying to reach out to Bountysource by email as per their new terms notification, and I'll make sure to make some noise if I don't get a prompt answer to make sure this can't get ignored. When the deployment is done, I'll make sure to add a message here, with instructions on how to collect the bounties when the issues get fixed.
Haiping
@Oceania2018
@jsboige @chriss2401 @henon Thanks.
I'm so exicted today, memory leak fixed and have a chance to do performance comparison, it's so awesome. https://twitter.com/Haiping2015/status/1276551033087557632
chriss2401
@chriss2401
@Oceania2018 that's awesome!!!
@jsboige did you consider this task : SciSharp/TensorFlow.NET#554
Jean-Sylvain Boige
@jsboige
Here are the new bounties:
https://www.bountysource.com/issues/91873084-inputs-was-changed-by-accidently
https://www.bountysource.com/issues/91873276-loading-of-hdf5-files-from-keras-not-possible
https://www.bountysource.com/issues/91873312-build-graph-implement-switch-statement-tests
https://www.bountysource.com/issues/78645236-implement-tf-image
@Oceania2018 congrats for the latest perfs. Since you said that issue was hard and important, I had half the available amount redeployed to the issue you indicated. I hope that works for you.
@chriss2401 I added your issue to the list. It's a good one indeed.
In order to collect the bounty, when the issue is closed, I believe you have to click on "Solve issue", and after a 2 weeks verification period you should be able to cash out with a 10% fee or redeploy at no fee.
Haiping
@Oceania2018
@jsboige Great Job!!!
Sam Harwell
@sharwell
@Oceania2018 Now I'm here :smile:
Haiping
@Oceania2018
@sharwell Welcome Sam, appreicate your work regarding to the SafeHandle. It helps us a lot.
emilekm2142
@emilekm2142
How to do cross product like in python with np.cross()?
There is no NumSharp.cross or <NDArray>.cross :cry:
emilekm2142
@emilekm2142
I have only NumSharp installed btw
chriss2401
@chriss2401
@emilekm2142 maybe try Numpy. NET instead
Meinrad Recheis
@henon
@emilekm2142 https://github.com/SciSharp/Numpy.NET has almost everything
tondangerza
@tondangerza
Hi, I train custom object detection (yolov3) in SciSharp/SciSharp-Stack-Examples with tensorflow.net 0.20 -peview-2 but line 101 in yolov3.cs (var conv = tf.nn.conv2d(input: input_layer, filter: weight, strides: strides, padding: padding);) is error Object reference not set to an instance of an object.' please can you help me?
Haiping
@Oceania2018
@tondangerza It's WIP. not ready yet.
tondangerza
@tondangerza
@Oceania2018 So, I must use tensorflow.net v1.50 and Tensorflow.Redist v.2.2.0.2 right?
Haiping
@Oceania2018
Both of them don't work for yolov3. The example hasn't been finished yet.
Amol Gholap
@amolgholap_twitter
Hi There, This is my first post and not sure which channel would be more appropriate so posting it here. Please guide me if it doesn't belong here. I wanted develop a service for hand track(hand raise) using TensorFlow.Net, are there any similar samples that you can point me to. Thanks for your help.
Haiping
@Oceania2018
@amolgholap_twitter Definitly, you are in the right place. you can capture the video frame, label the images, train a model. Check the examples: https://github.com/SciSharp/SciSharp-Stack-Examples , this is another computer vision tool for you: https://github.com/SciSharp/SharpCV
rootdarkarchon
@rootdarkarchon
Good day, is there a build for tensorflow.redist-windows-gpu 0.20 or should I take the 1.15.1?
Haiping
@Oceania2018
@rootdarkarchon Not yet.
Meinrad Recheis
@henon
@/all Numpy.NET finally has got complex number support
Haiping
@Oceania2018
@henon Congrats.
chriss2401
@chriss2401
@Oceania2018 will preview3 (or 4) include loading pre trained .h5 keras models and running inference on them?
Ø
Oh, and great job by the way!
Haiping
@Oceania2018
@chriss2401 no plan for .h5 yet.
Haiping
@Oceania2018
@/all .NET for TensorFlow released Preview3 version, which partially supports keras, dataset and eager of tensorflow 2.x, and supports the construction of neural network models. Download the NuGet installation package: https://www.nuget.org/packages/TensorFlow.NET/0.20.0-preview3, the most complete .NET AI tutorial is updated: https://github.com/SciSharp/TensorFlow.NET-Tutorials. Sample code: https://github.com/SciSharp/SciSharp-Stack-Examples, core code: https://github.com/SciSharp/TensorFlow.NET. Any feedback is welcome. Thank you!