sync dmlc/tvm 20190328

CONTRIBUTORS.md

@@ -1,10 +1,21 @@
 TVM Contributors
 ================
-TVM adopts the Apache style model and governs by merit. We believe that it is important to create an inclusive community where everyone can use,
+TVM adopts the Apache way and governs by merit. We believe that it is important to create an inclusive community where everyone can use,
 contribute to, and influence the direction of the project. We actively invite contributors who have earned the merit to be part of the development community.
 
 See the [community structure document](http://docs.tvm.ai/contribute/community.html) for the explanation of community structure and contribution guidelines.
 
+## Mentors
+
+TVM is now part of the Apache Incubator.
+We are fortunate to have the following mentors.
+
+- Markus Weimer @markusweimer
+- Sebastian Schelter @sscdotopen
+- Byung-Gon Chun @bgchun
+- Henry Saputra @hsaputra
+- Timothy Chen @tnachen
+- Furkan KAMACI @kamaci
 
 ## Committers
 

Jenkinsfile

@@ -22,7 +22,7 @@
 //
 ci_lint = "tvmai/ci-lint:v0.50"
 ci_gpu = "tvmai/ci-gpu:v0.51"
-ci_cpu = "tvmai/ci-cpu:v0.41"
+ci_cpu = "tvmai/ci-cpu:v0.50"
 ci_i386 = "tvmai/ci-i386:v0.50"
 
 // tvm libraries

apps/howto_deploy/Makefile

@@ -31,4 +31,4 @@ lib/cpp_deploy_pack: cpp_deploy.cc lib/test_addone_sys.o lib/libtvm_runtime_pack
 # Deploy using pre-built libtvm_runtime.so
 lib/cpp_deploy_normal: cpp_deploy.cc lib/test_addone_sys.o
 	@mkdir -p $(@D)
-	$(CXX) $(PKG_CFLAGS) -o $@  $^ $(PKG_LDFLAGS) -ltvm_runtime
+	$(CXX) $(PKG_CFLAGS) -o $@  $^ -ltvm_runtime $(PKG_LDFLAGS)

apps/howto_deploy/run_example.sh

@@ -3,8 +3,8 @@ echo "Build the libraries.."
 mkdir -p lib
 make
 echo "Run the example"
-export LD_LIBRARY_PATH=../../lib:${LD_LIBRARY_PATH}
-export DYLD_LIBRARY_PATH=../../lib:${DYLD_LIBRARY_PATH}
+export LD_LIBRARY_PATH=../../build:${LD_LIBRARY_PATH}
+export DYLD_LIBRARY_PATH=../../build:${DYLD_LIBRARY_PATH}
 
 echo "Run the deployment with all in one packed library..."
 lib/cpp_deploy_pack

cmake/config.cmake

@@ -127,7 +127,7 @@ set(USE_MPS OFF)
 set(USE_ROCBLAS OFF)
 
 # Whether use contrib sort
-set(USE_SORT OFF)
+set(USE_SORT ON)
 
 # Whether use TensorRT
 # /path/to/tensorrt that contains include and lib dirs

cmake/modules/contrib/BLAS.cmake

@@ -10,7 +10,7 @@ elseif(USE_BLAS STREQUAL "mkl")
   if(NOT IS_DIRECTORY ${USE_MKL_PATH})
     set(USE_MKL_PATH /opt/intel/mkl)
   endif()
-  find_library(BLAS_LIBRARY mkl_rt ${USE_MKL_PATH}/lib/ ${USE_MKL_PATH}/lib/intel64)
+  find_library(BLAS_LIBRARY NAMES mkl_rt mklml_gnu HINTS ${USE_MKL_PATH}/lib/ ${USE_MKL_PATH}/lib/intel64)
   include_directories(${USE_MKL_PATH}/include)
   list(APPEND TVM_RUNTIME_LINKER_LIBS ${BLAS_LIBRARY})
   list(APPEND RUNTIME_SRCS ${CBLAS_CONTRIB_SRC})

dmlc_tvm_commit_id

@@ -1 +1 @@
-5f37cc1290355c840e36ba3a4a7ff158a97896c5
+a0537ecbf80fafee22ea64bb2994332d4488b81f

include/tvm/relay/attrs/nn.h

@@ -155,6 +155,24 @@ struct Conv2DWinogradAttrs : public tvm::AttrsNode<Conv2DWinogradAttrs> {
   }
 };
 
+/*! \brief Attributes used in winograd weight transformation operators */
+struct Conv2DWinogradNNPACKWeightTransformAttrs
+    : public tvm::AttrsNode<Conv2DWinogradNNPACKWeightTransformAttrs> {
+  int convolution_algorithm;
+  DataType out_dtype;
+
+  TVM_DECLARE_ATTRS(Conv2DWinogradNNPACKWeightTransformAttrs,
+                    "relay.attrs.Conv2DWinogradNNPACKWeightTransformAttrs") {
+    TVM_ATTR_FIELD(convolution_algorithm)
+        .describe(
+            "The convolution algorithm for Winograd NNPACK. "
+            "E.g. tvm.contrib.nnpack.ConvolutionAlgorithm.WT_8x8 for WT_8x8, "
+            "tvm.contrib.nnpack.ConvolutionAlgorithm.WT_8x8_FP16 for WT_8x8_FP16");
+    TVM_ATTR_FIELD(out_dtype)
+        .set_default(NullValue<DataType>())
+        .describe("Output data type, set to explicit type under mixed precision setting");
+  }
+};
 
 /*! \brief Attributes used in softmax operators */
 struct SoftmaxAttrs : public tvm::AttrsNode<SoftmaxAttrs> {

nnvm/include/nnvm/top/nn.h

@@ -183,6 +183,26 @@ struct WinogradWeightTransformParam : public dmlc::Parameter<WinogradWeightTrans
     static const constexpr int kWeight = 0;
 };
 
+struct WinogradNNPACKWeightTransformParam
+    : public dmlc::Parameter<WinogradNNPACKWeightTransformParam> {
+  int convolution_algorithm;
+  int out_dtype;
+
+  DMLC_DECLARE_PARAMETER(WinogradNNPACKWeightTransformParam) {
+    DMLC_DECLARE_FIELD(convolution_algorithm)
+        .describe(
+            "The convolution algorithm for Winograd NNPACK. "
+            "E.g. tvm.contrib.nnpack.ConvolutionAlgorithm.WT_8x8 for WT_8x8, "
+            "tvm.contrib.nnpack.ConvolutionAlgorithm.WT_8x8_FP16 for WT_8x8_FP16");
+    DMLC_DECLARE_DTYPE_FIELD(out_dtype)
+        .add_enum("same", -1)
+        .set_default(-1)
+        .describe("Output data type, set to explicit type under mixed precision setting");
+  }
+
+  static const constexpr int kWeight = 0;
+};
+
 struct WinogradConv2DParam : public dmlc::Parameter<WinogradConv2DParam> {
   int channels;
   TShape kernel_size;

nnvm/python/nnvm/frontend/caffe2.py

@@ -3,7 +3,7 @@
 from __future__ import absolute_import as _abs
 import tvm
 from nnvm import symbol as _sym
-from nnvm.frontend.common import get_nnvm_op, Renamer, AttrConverter as AttrCvt
+from .common import get_nnvm_op
 from .onnx_caffe2_utils import dimension_picker, dimension_constraint, infer_channels, revert_caffe2_pad
 from . import onnx
 
@@ -73,8 +73,8 @@ def get_converter(cls):
 
         if hasattr(cls, '_impl'):
             return getattr(cls, '_impl')
-        raise NotImplementedError('{} not implemented'.format(
-            cls.__name__))
+        raise tvm.error.OpNotImplemented(
+            'Operator {} is not implemented in frontend Caffe2.'.format(cls.__name__))
 
 
 _caffe2_internal_args = {
@@ -176,8 +176,7 @@ def _get_axis_from_order_str(order):
                 return 1
             if order == 'NHWC':
                 return 3
-            raise RuntimeError(
-                "Unsupported storage order: {} in caffe2".format(order))
+            raise tvm.error.OpAttributeInvalid('Value {} in attribute {} of operator {} is not valid.'.format(order, 'order', 'Concat'))
 
         return AttrCvt(
             op_name='concatenate',
@@ -427,8 +426,8 @@ def _convert_operator(self,
             # Add a sanitizing step to convert all byte strings in args to strings
             sym = convert_map[op_type](inputs, args, self._params)
         else:
-            raise NotImplementedError(
-                "Operator {} not implemented.".format(op_type))
+            raise tvm.error.OpNotImplemented(
+                'Operator {} is not supported in frontend Caffe2.'.format(op_type))
         return sym
 
 

nnvm/python/nnvm/frontend/common.py

@@ -7,9 +7,25 @@
 def get_nnvm_op(op_name):
     op = getattr(_sym, op_name)
     if not op:
-        raise RuntimeError("Unable to map op_name {} to nnvm.sym".format(op_name))
+        raise OpNotImplemented(
+            'Operator {} is not supported.'.format(op))
     return op
 
+def required_attr(attr, key, op_name):
+    assert isinstance(attr, dict)
+    if key not in attr:
+        raise OpAttributeRequired(
+            'Required attribute {} not found in operator {}'.format(key, op_name))
+    return attr[key]
+
+def parse_tshape(tshape):
+    """Parse tshape in string."""
+    return [int(x.strip()) for x in tshape.strip('()').split(',')]
+
+def parse_bool_str(attr, key, default='False'):
+    """Parse bool string to boolean."""
+    return attr.get(key, default).strip().lower() in ['true', '1', 't', 'y', 'yes']
+
 class Renamer(object):
     """A simply renamer for operators.
 

nnvm/python/nnvm/frontend/coreml.py

@@ -2,11 +2,10 @@
 """CoreML frontend."""
 from __future__ import absolute_import as _abs
 import numpy as np
-
 import tvm
+from .common import SymbolTable
 from .. import symbol as _sym
 from .._base import string_types
-from .common import SymbolTable
 
 __all__ = ['from_coreml']
 
@@ -83,7 +82,8 @@ def BatchnormLayerParams(op, insym, symtab):
     """Get layer of batchnorm parameter"""
     # this changes the symbol
     if op.instanceNormalization:
-        raise NotImplementedError("instance normalization not implemented")
+        msg = 'Operator "instance normalization" is not supported in frontend CoreML.'
+        raise tvm.error.OpNotImplemented(msg)
     else:
         params = {'gamma':symtab.new_const(list(op.gamma.floatValue)),
                   'beta':symtab.new_const(list(op.beta.floatValue)),
@@ -136,7 +136,8 @@ def ActivationParams(op, insym, symtab):
         betasym = symtab.new_const(beta)
         return _sym.broadcast_mul(_sym.log(_sym.broadcast_add(
             _sym.exp(insym), betasym)), alphasym)
-    raise NotImplementedError('%s not implemented' % whichActivation)
+    raise tvm.error.OpNotImplemented(
+        'Operator {} is not supported in frontend CoreML.'.format(whichActivation))
 
 def ScaleLayerParams(op, insym, symtab):
     """Scale layer params."""
@@ -158,7 +159,8 @@ def PoolingLayerParams(op, insym, symtab):
             return _sym.global_max_pool2d(insym)
         if op.type == 1:
             return _sym.global_avg_pool2d(insym)
-        raise NotImplementedError("Only max and average pooling implemented")
+        raise tvm.error.OpNotImplemented(
+            'Operator pooling (not max or average) is not supported in frontend CoreML.')
 
     else:
         params = {'pool_size':list(op.kernelSize),
@@ -178,7 +180,8 @@ def PoolingLayerParams(op, insym, symtab):
             params['padding'] = padding
             params['ceil_mode'] = True
         else:
-            raise NotImplementedError("Other convolution padding not implemented")
+            msg = 'Value {} in attribute PoolingPaddingType of operator Pooling is not valid.'
+            raise tvm.error.OpAttributeInvalid(msg.format(op.WhichOneof('PoolingPaddingType')))
 
         # consume padding layer
         if symtab.in_padding:
@@ -190,7 +193,8 @@ def PoolingLayerParams(op, insym, symtab):
             return _sym.max_pool2d(insym, **params)
         if op.type == 1:
             return _sym.avg_pool2d(insym, **params)
-        raise NotImplementedError("Only max and average pooling implemented")
+        msg = 'Operator pooling (not max or average) is not supported in frontend CoreML.'
+        raise tvm.error.OpNotImplemented(msg)
 
 def SoftmaxLayerParams(op, insym, symtab):
     return _sym.softmax(_sym.flatten(insym))
@@ -229,7 +233,8 @@ def ConcatLayerParams(op, insyms, symtab):
     if not isinstance(insyms, list):
         insyms = [insyms]
     if op.sequenceConcat:
-        raise NotImplementedError("Sequence Concat not supported")
+        raise tvm.error.OpNotImplemented(
+            'Operator Sequence Concat is not supported in frontend CoreML.')
     ret = _sym.concatenate(*insyms, axis=1)
     return ret
 
@@ -243,14 +248,16 @@ def PaddingLayerParams(op, insym, symtab):
     if op.WhichOneof('PaddingType') == 'constant':
         constant = op.constant
         if constant.value != 0:
-            raise NotImplementedError("Padding value {} not supported.".format(constant.value))
+            msg = 'Value {} in attribute "padding value" of operator Padding is not valid.'
+            raise tvm.error.OpAttributeInvalid(msg.format(constant.value))
         padding = [b.startEdgeSize for b in op.paddingAmounts.borderAmounts]
         padding2 = [b.endEdgeSize for b in op.paddingAmounts.borderAmounts]
         for i, j in zip(padding, padding2):
             assert i == j
         symtab.set_padding(padding)
     else:
-        raise NotImplementedError("Only constant padding is supported now.")
+        raise tvm.error.OpNotImplemented(
+            'Operator "non-constant padding" is not supported in frontend CoreML.')
     return insym
 
 def PermuteLayerParams(op, insym, symtab):
@@ -259,8 +266,8 @@ def PermuteLayerParams(op, insym, symtab):
 
 def UpsampleLayerParams(op, insym, symtab):
     if op.scalingFactor[0] != op.scalingFactor[1]:
-        raise NotImplementedError("Upsampling only supported with same \
-            height and width scaling factor.")
+        raise tvm.error.OpAttributeInvalid(
+            'Height and width scaling factors of Upsample operator must be equal.')
     interpolationMode = 'NEAREST_NEIGHBOR' if op.mode == 0 else 'BILINEAR'
     return _sym.upsampling(insym, scale=op.scalingFactor[0], method=interpolationMode)
 
@@ -341,7 +348,8 @@ def coreml_op_to_nnvm(op, inname, outname, symtab):
     """
     classname = type(op).__name__
     if classname not in _convert_map:
-        raise NotImplementedError("%s is not supported" % (classname))
+        raise tvm.error.OpNotImplemented(
+            'Operator {} is not supported in frontend CoreML.'.format(classname))
     if isinstance(inname, string_types):
         insym = symtab.get_var(inname)
     else: