1. 11 Jul, 2019 4 commits
    • andreykhaykin's avatar
      Tracer of real data in ngraph-CPU for debug purpose (#3099) · 4b009f09
      andreykhaykin authored
      * new debugger to trace real tensors in each kernel
      
      * update after style-apply
      
      * add unit test to debugger
      
      * after style-apply
      
      * templatize data + fixed issues from PR
      
      * solve conflict
      
      * change names of env flags + fixed typo + pass unit-test
      
      * add different types to dump, change arch of debug class, add support to allow debug to be set programmatically (all requested changed fixed)
      
      * add curly braces to single-line blocks
      
      * fix shadow names + add int32 for dump + fix for names with coding standarts of ngraph
      
      * fix names of member variables + delete some constructors
      
      * try to find issue with CI, do not merge!
      
      * fix one issue with CI
      
      * fix style-apply
      4b009f09
    • Nagy Mostafa's avatar
      [MLIR] Support MLIR lowering of Relu (#3197) · d1af0bb7
      Nagy Mostafa authored
      * Support MLIR lowering of Relu
      
      * Use EDSC comparison
      
      * style-apply
      
      * Use .inc file for Conversion classes list
      
      * Disable i32 Relu for plaidml
      d1af0bb7
    • Rob Earhart's avatar
      Use libc++ for macOS core python wheel (#3177) · a655e1cc
      Rob Earhart authored
      * Use libc++ for macOS core python wheel
      
      * numpy version
      a655e1cc
    • Adam Procter's avatar
      Enable ConstantFolding for Product, Reverse; add a couple of dynamic tests (#3187) · 0611ae23
      Adam Procter authored
      * Add a few more toy dynamic examples
      
      * Enable constant folding for Reverse; enable reverse_shape test
      
      * Add constant folding for Product, and enable two more tests
      0611ae23
  2. 10 Jul, 2019 9 commits
  3. 09 Jul, 2019 16 commits
  4. 08 Jul, 2019 5 commits
  5. 06 Jul, 2019 2 commits
  6. 04 Jul, 2019 3 commits
    • Rob Earhart's avatar
      Accept empty PlaidML config options · a18ad791
      Rob Earhart authored
      a18ad791
    • Rob Earhart's avatar
      Replace get_arguments().at(x) with get_argument(x) · adddbaa1
      Rob Earhart authored
      This came up while debugging memory corruption in the ngraph-tf tests: plaidml_ops_replicate was holding onto
      a reference to a temporary that was destroyed while the reference was still live.  Replacing that code with a
      call to get_argument() eliminated the temporary and fixed the issue.  And in general, there's no reason to be
      allocating temporary vectors when we're pulling out a single argument, so this change replaces the other
      sequences of get_arguments().at(), too.
      adddbaa1
    • Rob Earhart's avatar
      ReshapeElimination -> PlaidML; fixup ImplicitBcast · 1144bf39
      Rob Earhart authored
      The PrefixReshapeElimination pass doesn't work correctly if the network's operation's input and output shape
      compatibility is being validated; an implicit broadcast operation is required to get the shapes right.  This
      change moves the implementation back to PlaidML (since it's seeming less generally useful, and PlaidML has an
      implicit broadcast operation), and fixes both it and the implicit broadcast pass to only apply when the
      downstream operation is an elementwise operation (since PlaidML doesn't support automatic broadcasting for
      contractions).
      1144bf39
  7. 03 Jul, 2019 1 commit