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

Converters between org.nd4j.linalg.primitives.Pair/Triple and scala Tuple #1

Open
wants to merge 32 commits into
base: master
Choose a base branch
from

Conversation

zorba128
Copy link
Owner

No description provided.

Yurii and others added 30 commits August 31, 2018 00:15
* fixing and testing split op

* add to space_to_batch op possiblity to work with blocks of length=1
…uple2/Tuple3.

Names to be decided - as proposed, or maybe toScala or toTuple
* initial commit

* one more commit

* update 1.9

* tmp

* pew

* cmake

* cmake

* cmake

* draft

* next commit

* next commit

* next commit

* typo

* schema update

* one more commit

* configurable port

* meh

* next commit

* next commit

* separate cmake

* few tweaks

* typo

* few more tweaks

* dd

* get rid of old conf

* avx/avx2 default

* next commit

* typo

* const

* next commit

* Dockerfile draft

* Dockerfile draft 2.0

* Dockerfile draft 2.1

* - Dockerfile draft 3.0
- first test passes

* cmake again

* cmake again

* copyback

* slash

* releasable result

* - one more test
- Graph::printOut now includes variables

* minor tweak

* scema update 1

* schema update 2

* schema update 3

* schema update 2

* schema update 4

* schema update 2

* schema update 6

* one more test

* VariableProxy used now

* filename/port

* const fix

* T

* port number

* bindings update on java side

* jcpp compat fix

* grpc stuff as separate module

* GraphInferenceClient prototype

* commit to trigger jenkins

* GraphInferenceClient prototype

* update

* few UX tweaks

* one test

* one test

* one test

* schema change

* schema update

* schema change

* schema change

* expand dims

* a bit more logging

* fetch orginal ids by string

* prototype works

* some error handling

* schema update

* schema update

* RWL

* ...

* next step

* error handling

* typo

* typo

* typo

* typo

* typo

* just a float

* execute impl

* lol

* initial values

* one more exception

* one more test

* some logging

* missed include

* tests tweaks

* - Dockerfile updates
- README.md

* MD

* Dockerfile label

* readme update

* error handling

* flatbuffers update

* doc update

* tests tweaked

* one more tweak

* one more tweak

* disable tests by default
* Fixed 1D bug with in_top_k op.

* Fixed wrong char error.
* matrix_band_part op. Initial version.

* Added matrix_band_part op and tests.

* Added helpers for matrix_band_part op.
…6310)

* deeplearning4j#6305 Nd4j.where empty arrays

* Fix SplitV op + TF import

* Fix TF import shape issue

* Small SplitV fix

* Map SparseSoftmaxCrossEntroyWithLogits op

* Mapping fix for sparse softmax loss

* Add TF mapping for Where op

* BroadcastDynamicShape import mapping; re-enable now fixed ops

* Re-enable new tests

* Switch Where op for import from where_np to where

* Tweaks before merge

* A few extra ignores for new tests
* deeplearning4j#6327 INDArray.put with SpecifiedIndex

* deeplearning4j#6341 - SpecifiedIndex with single value no longer collapses dimensions

* Another indexing fix

* deeplearning4j#6343 TransferLearning nOutReplace fix
* implementation and testing broadcast_to op

* correct tile_to_shape op

* correct tests for tile_to_shape op
* Shugeo dynamic bp (deeplearning4j#6106)

* Added dynamic_partition_bp op and helpers. Initial revision.

* Implementation of dynamic_partion_bp and test.

* Fixed implementation bugs.

* Minor fixes with dynamic_partition_bp and test.

* Modified dynamic_partition_bp op and tests.

* Modified dynamic_stitch invocation with dynamic_partition_bp op.

* Another approach to implement.

* Final version of dynamic_partition_bp op.

* Refactored revision of dynamic_partition_bp op.

* The first working revision.

* initial commit

* another initial commit

* another initial commit

* yet another initial commit

* yet another initial commit

* yet another initial commit

* one more initial commit

* one more initial commit

* equality

* moar equality

* serialization

* serialization

* skeleton draft

* one more step

* next step

* next step

* minor rearrangements

* next step

* next step

* time to flow

* next step

* one more step

* more structure for skeleton

* temp commit

* one step closer

* time to tests stuff

* one more step

* one more step

* few more assertions

* one failing test

* propagation direction

* one small step

* another small step

* test passes now

* in-memory chunks tracker

* reconnect after failure test

* few callbacks

* temp commit

* commit for alex

* remove original hook for few messages

* first MPS test

* one more step

* gradients update propagation

* ignored legacy tests

* one more step

* draft tracker switcher

* 2k nodes remap test

* delayed transport

* first few delayed tests

* first fixes to ensure safe mesh updates & propagation

* - proper nodes mapping into mesh
- serialization tweaks

* one more test

* temp commit

* one more commit

* plain mode added

* aeron version update

* aeron time

* still aeron time

* and aeron time again

* wow, aeron time!

* strange, but still aeron time!

* few rearrangements

* vew more rearrangements

* first test passes

* onMeshUpdate

* you can't touch this

* dl4j draft

* model/updater params propagation

* mesh build mode propagation

* few tweaks

* few more tweaks

* few more tweaks

* few more tweaks

* commit to trigger jenkins

* few more tweaks

* some more logging

* some more logging

* few missing constructor

* simple serde test

* explicit messages init

* :/

* one more serde test

* another deserializer

* lang3 SerializationUtils go

* some more logging

* micro tweak

* micro tweak

* few tests updated

* few more tweaks

* HeartbeatThread draft

* attach heartbeat

* temporary limited number of downstreams for Alex

* propagate model and updater params

* one failing test

* ping/pong stubs

* get rid of sync on addConnection()

* a bit more logging

* a bit more logging

* - bunch of tweaks
- couple of failing tests

* few tests tweaks

* special assertion

* updates handling

* messages thread

* two more tweaks

* a bit more logging

* shutdown hook

* consumer attached

* few things for debugging

* few things for debugging

* few things for debugging

* few things for debugging

* few things for debugging

* few things for debugging

* decrease number of nodes

* few things for debugging

* few failing tests

* mesh-related updates

* one more test for MESH mode

* remap test + tweaks

* version update

* temporary decrease number of downstreams per node

* some logging

* sync on start

* master always connected

* get rid of sleep in PW

* broadcast sender pulled off main thread

* exec pool size

* some deprecated annotations

* bunch of minor fixes

* bunch of minor fixes
* Fixed order bug with top_k op for unsorted cases.

* Fixed ugly error with test assert.
* Added a pair tests for Where op.

* Properly processed where on false-only case.

* Fixed only false case for where and where_np ops.

* Eliminate waste output with test.
* Fixed broadcast_dynamic_shape op 1D special case.

* Eliminate expensive tests.
* more builder options

* read config

* param server type

* tests
Fix IllegalArgumentException when inputs and labels have different sizes.
* Switch back to spark 1 as default

* Fix issue with Nd4j.where()

* Fix elephas import build issue
* usorted_segment_* ops implementation. Initial version.

* with proper unsorted_segment_sqrt_n helper signature.

* Implemented index check up for segment_* ops.

* 1D case for unsorted_segment_max was implemented and tested.

* Implemented and tested unsorted_segment_max op.

* Implemented and tested unsorted_segment_min op.

* Implemented and tested unsorted_segment_sum op.

* Implemented and tested unsorted_segment_prod op.

* Implemented unsorted_segment_mean and tests.

* Implemented and tested unsorted_segment_sqrt_n op.
* deeplearning4j#6358 Fix issue with LayerWorkspaceMgr when used directly in conjunction with CuDNN

* 6326 RnnOutputLayer + CompGraph + Dropout fix

* Another test + improve RnnOutputLayer error

* deeplearning4j#6368 Fix MultiLayerNetwork pretrainLayer iterator reset supported check

* deeplearning4j#6352 fix use of assert keywork for argument validation in ND4J

* deeplearning4j#6352 more assert keyword fixes

* deeplearning4j#6366 RecordReaderMultiDataSetIterator: better validation/errors; 5D (3D CNN) NDArrayWritable support

* Cleanup

* Compilation fix

* More cleanup

* Removed method fix

* More fixes
AlexDBlack and others added 2 commits September 8, 2018 13:24
* Arbiter test config fixes

* Some ParallelInference fixes + test fixes

* More test fixes
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

9 participants