Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

SameDiff: add java-level dtype check for ops #6861

Closed
AlexDBlack opened this issue Dec 15, 2018 · 1 comment

Comments

Projects
None yet
1 participant
@AlexDBlack
Copy link
Member

commented Dec 15, 2018

Consider the following:

SDVariable x = ...
SDVariable y = ...
SDVariable eq = x.eq(y);
SDVariable sum = eq.sum();

Note that variable eq has type BOOL. sum() op should fail for this case (as it does with INDArray.sum()), but doesn't here.

In general, we should check and immediately throw exceptions when ops are applied on SDVariables with an invalid datatype.

@AlexDBlack AlexDBlack added the SameDiff label Dec 15, 2018

@AlexDBlack AlexDBlack self-assigned this Apr 11, 2019

@AlexDBlack AlexDBlack added this to the Release Milestone milestone Apr 11, 2019

AlexDBlack added a commit that referenced this issue Apr 17, 2019

[WIP] QA, fixes, DL4J net convertDataType methods (#7531)
* Fix BaseNDArray.equalsWithEps issue for scalars of different ranks

* #7447 Fix slice on row vector

* #7483 Remove old deserialization warnings

* #6861 SameDiff datatype validation, round 1

* #6861 SameDiff datatype validation, round 2

* #6861 SameDiff datatype validation, round 3

* More rank 2 minimum shape fixes

* Multiple test fixes after changing rank2 minimum shapes

* Test fixes

* #7520 add MultiLayerNetwork.convertDataType(DataType) + test

* Datatype cleanup and fixes

* DL4J: Fixes for global (default) vs. network datatypes

* Fix incorrect datatype when arrays (different to default dtype) are detached

* Multiple fixes, improve tests

* Test

* #7532 New network datatype configuration

* Pass network dtype to layer/vertex initialization

* Yolo datatype fixes

* More fixes, more tests

* More fixes, more tests

* Fix bug in PoolHelperVertex backprop

* Vertex dtype tests; misc fixes

* Fix for BaseReduce3Op dtype

* More fix; finally all layers/vertices/preprocessors tested for dtypes

* Fix slices()

* Fixes - gradient check dtype issues

* Pass network dtype when constructing layers

* Pass network dtype when constructing vertices

* Layer dtype/casting fixes

* Various fixes

* Fix Shape.elementWiseStride for 1d view case

* #7092 INDArray.get(point,x)/get(x,point) returns 1d array

* More 1d getRow/getCol fixes

* Indexing/sub-array fixes

* More test and indexing fixes

* More test fixes, add getRow(i,keepDim) and getColumn(i,keepDim)

* More indexing/test fixes

* More fixes

* More fixes

* More fixes

* #7550 Evaluation dtype tests + fixes

* Nd4j.gemm result dtype fix

* Next round of fixes

* Even more dtype fixes...

* Datavec and more DL4J fixes

* Next round of fixes

* DL4J cuDNN helpers - dtype improvements/fixes

* Another round of fixes

* Datavec fixes

* DL4J Fixes

* Keras/Spark/elementwisevertex fixes

* Final (hopefully) fixes

* Last set of fixes
@lock

This comment has been minimized.

Copy link

commented May 17, 2019

This thread has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs.

@lock lock bot locked and limited conversation to collaborators May 17, 2019

Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
You can’t perform that action at this time.