rahular/darviz-issues

This repository is meant for raising and monitoring issues related to @ibmdarviz

darviz-issues's Issues

I can't go to the website of dlide.mybluemix.net

Is it some problem with the website? https://dlide.mybluemix.net/

Initial issue from GUI

Hope it works :)

Submitted by: [email protected]

from directly gmail adding issue

please ignore.

delete

I am playing around with the interface, but I can't figure out how to delete a component. Because it's an alpha version, I'm not sure if it's not implemented or I'm just missing it.

Is there a way to generate (Keras) code form a partially completed model?

If a model is for instance temporarily lacking a data source or that source is in some other format than Pickle and one wishes to fill in code in Keras is there a way to generate code from a prtially completed design?

VGG 19

First of all..

Thank you !!!!!!! for building such an awesome ide to make life simple...

Why aren't the original model with the dropout layers and all is there?(original is original)

like this one

Also after generating the code from the ide it will be much better if the layer retains their name as
implied by the layer..

this

ReLU_2 = Activation('relu', name = 'ReLU.7e9f24ad')(Convolution2D_1)

to

ReLU_2 = Activation('relu', name = 'ReLU_2')(Convolution2D_1) (removed the hash_id)

I know its not necessary but it will help in tensorboard while visualisation..

Initial issue

This is a test issue

Superresolution

If I am correct, DARVIZ is used for classification tasks.
How to use this tool for a super-resolution task?

Cannot import name SigmoidCrossEntropyLoss

Hi, when i tried to run the MNIST tutorial from TensorFlow, I had this issue " File "Keras.py", line 14, in
from keras.layers import SigmoidCrossEntropyLoss, Activation, Input
ImportError: cannot import name SigmoidCrossEntropyLoss"

ImageData nodes cannot be saved, Create SourceCode does not work

Error1:
Every time I try to save an ImageData node and use a "Pre-exisiting" model (yes that typo is there as well...) such as MINIST, CIFAR10 or others, the node will NOT be saved but it will switch back to Datasource: "Create New".
So I cannot run any model.

Error2:
Also when I try to "Create Source Code" it just tells me to
"Please create a design model to be able to create a job." (well I already did before...)

Question:
Also if I export a model to a model.dz I get a file but I have no idea what to do with it from there. What do I do with it?

Please OPEN SOURCE DARVIZ IDE on github so that we can contribute , extend it and develop it further

Please Open Source DARVIZ IDE on github so that we can contribute , extend it and develop it further. Since it runs on Node-RED and Node.js and currently uses Keras to support /wrap/abstract Tensorflow, Theano and Caffe as a backend. We love it but without it being Open Sourced then very few developers will participate and such novel idea will just die . I don't want it to die but i want to promote it because it brings developers together to extend it. I am Tinkerer and i like to develop and code various AI models but i believe we can quickly extend and contribute to Darviz IDE for a much Larger adoption.

Google made a smart move and out sourced Tensorflow and look how quickly it was adopted by developers and list goes on, Keras , Caffe , MXNet , and others.

The Mantra for developers in this century is colloboration using Open Source. or risk a good idea just dying very quickly .

How do I delete or edit nodes in Designer?

I mistakenly created a node I don't wish to incorporate into the model how do I delete it?

Cannot find input layers for csv files.

Unable to load data model

Hello sir, i want to load the data model but it is not showing.

Submitted by: [email protected]

Compare & Contrast Darviz with Barista https://www.uni-muenster.de/PRIA/Barista/

As a graphical tool for Deep Learning Could You Compare the Features and Limitations of Darviz vs. the Barista project for Cafee out of Muenster

darviz public git with slack testing

please ignore

Pytorch

Do you have plans to support pytorch. It might be a good idea as Pytorch is gaining popularity in DL very rapidly.

Consider the Use Case Where You Wish to Use Darviz to Design NN for Own Data

In preparation for using Darviz as a design tool for NN's to be processed in Keras/TF on my own data sets I generated Keras Code with the MNIST Example from your site with Model A. I then loaded it to my Ubuntu Docker Instance of Keras/TF and attempted to edit the code to take the MNIST image from Keras itself.
I would like to use Darviz as a tool to design efficient NN's but incorporate calculations on my own data sets. What would you recommend as the best approach and how can I best get Keras code to be edited for my own data sets and formats stored on the machine running compilation to Tensorflow?

darviz-git plus bluemix testing

Please Ignore

Submitted by: [email protected]

test bug from bluemix

test

Submitted by: [email protected]

How does one delete a design from the Menu.Storage Area?

New data model is not coming

my own old data model loading problem

Darviz does not generate the code

I am trying since last 2 days it allows to create the model but does not generate the code.