[incubator-mxnet] branch master updated (24f0a10 -> 4f8bc3a)
This is an automated email from the ASF dual-hosted git repository. haibin pushed a change to branch master in repository https://gitbox.apache.org/repos/asf/incubator-mxnet.git. from 24f0a10 [MXNET-978] Higher Order Gradient Support `clip`, `dropout`. (#15746) add 4f8bc3a Fix unary operator ceil/floor/trunc when data type is integer (#14251) No new revisions were added by this update. Summary of changes: src/operator/math_functions-inl.h | 15 ++- tests/python/unittest/test_ndarray.py | 19 +++ 2 files changed, 29 insertions(+), 5 deletions(-)
[GitHub] [incubator-mxnet] eric-haibin-lin closed issue #13220: Many operators don't work for integer type
eric-haibin-lin closed issue #13220: Many operators don't work for integer type URL: https://github.com/apache/incubator-mxnet/issues/13220 This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] [incubator-mxnet] eric-haibin-lin merged pull request #14251: Fix unary operator ceil/floor/trunc when data type is integer
eric-haibin-lin merged pull request #14251: Fix unary operator ceil/floor/trunc when data type is integer URL: https://github.com/apache/incubator-mxnet/pull/14251 This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] [incubator-mxnet] eric-haibin-lin commented on issue #14251: Fix unary operator ceil/floor/trunc when data type is integer
eric-haibin-lin commented on issue #14251: Fix unary operator ceil/floor/trunc when data type is integer URL: https://github.com/apache/incubator-mxnet/pull/14251#issuecomment-529075325 thx for the fix This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] [incubator-mxnet] sxjscience commented on issue #16102: Usability degradation
sxjscience commented on issue #16102: Usability degradation URL: https://github.com/apache/incubator-mxnet/issues/16102#issuecomment-529074657 @anirudh2290 Have we tested the case for reshaping to an invalid shape? This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] [incubator-mxnet] samskalicky commented on a change in pull request #15921: [WIP] dynamic custom operator support
samskalicky commented on a change in pull request #15921: [WIP] dynamic custom
operator support
URL: https://github.com/apache/incubator-mxnet/pull/15921#discussion_r321955499
##
File path: example/lib_ops/subgraph_lib.cc
##
@@ -0,0 +1,125 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements. See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership. The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License. You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing,
+ * software distributed under the License is distributed on an
+ * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+ * KIND, either express or implied. See the License for the
+ * specific language governing permissions and limitations
+ * under the License.
+ */
+
+/*!
+ * Copyright (c) 2019 by Contributors
+ * \file subgraph_lib.cc
+ * \brief subgraph operator implementation
+ * library file
+ */
+
+#include
+#include "lib_api.h"
+
+MXReturnValue parseAttrs(std::map attrs,
+ int* num_in, int* num_out) {
+ *num_in = 2;
+ *num_out = 1;
+ return MX_SUCCESS;
+}
+
+MXReturnValue inferType(std::map attrs,
std::vector ,
+ std::vector ) {
+ outtypes[0] = intypes[0];
+ return MX_SUCCESS;
+}
+
+MXReturnValue inferShape(std::map attrs,
std::vector> ,
+ std::vector> ) {
+ unsigned n = inshapes[0][0];
+ unsigned k = inshapes[0][1];
+ unsigned kk = inshapes[1][0];
+ unsigned m = inshapes[1][1];
+
+ std::cout << "inshapes[0][0]=" << n << " inshapes[0][1]=" << k << std::endl;
+ std::cout << "inshapes[1][0]=" << kk << " inshapes[1][1]=" << m <<
std::endl;
+
+ if (k != kk)
+return MX_FAIL;
+
+ outshapes[0].push_back(n);
+ outshapes[0].push_back(m);
+ return MX_SUCCESS;
+}
+
+MXReturnValue mutateInputs(std::map attrs,
+ std::vector _indices) {
+ input_indices.push_back(1);
+ std::cout << "the 1st input is marked as mutate input by library author" <<
std::endl;
+ return MX_SUCCESS;
+}
+
+class MyStatefulOp : public CustomStatefulOp {
+ public:
+ MyStatefulOp(std::string sym, int count) : subgraph_sym(sym), count(count) {}
+
+ void Forward() {
+count++;
+ }
+
+ int State() {
+return count;
+ }
+
+ ~MyStatefulOp() {}
+
+ private:
+ std::string subgraph_sym;
+ int count;
+};
+
+MXReturnValue createOpState(std::map attrs,
+CustomStatefulOp** op_inst) {
+ *op_inst = new MyStatefulOp("json", 0);
Review comment:
Lets add a check for "subgraph_sym_json" key in the attrs map, and error out
if its missing.
Lets get the "json" string from the attrs map using the "subgraph_sym_json"
key rather than setting the placeholder "json" string.
We'll assume the subgraph property sets this attr, later.
```
CHECK(attrs.find("subgraph_sym_json") != attrs.end()) << "Error! Expected
key '" << "subgraph_sym_json" << "' in Node Attributes";
*op_inst = new MyStatefulOp(attrs["subgraph_sym_json"], 0);
```
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
With regards,
Apache Git Services
[GitHub] [incubator-mxnet] samskalicky commented on a change in pull request #15921: [WIP] dynamic custom operator support
samskalicky commented on a change in pull request #15921: [WIP] dynamic custom
operator support
URL: https://github.com/apache/incubator-mxnet/pull/15921#discussion_r321955177
##
File path: include/mxnet/lib_api.h
##
@@ -18,33 +18,627 @@
*/
/*!
- * Copyright (c) 2015 by Contributors
+ * Copyright (c) 2019 by Contributors
* \file lib_api.h
* \brief APIs to interact with libraries
+ * This API specifies function prototypes to
+ * register custom ops for library authors
*/
+
#ifndef MXNET_LIB_API_H_
#define MXNET_LIB_API_H_
+#include
+#include
+#include
+#include
+
+#define MX_LIBRARY_VERSION 1
+
/*!
- * \brief Following are the APIs implemented in the external library
+ * \brief External Tensor data types
+ */
+enum MXDType {
+ kFloat32 = 0,
+ kFloat64 = 1,
+ kFloat16 = 2,
+ kUint8 = 3,
+ kInt32 = 4,
+ kInt8 = 5,
+ kInt64 = 6,
+};
+
+enum MXReturnValue {
+ MX_FAIL = 0,
+ MX_SUCCESS = 1,
+};
+
+/*!
+ * \brief External Tensor data structure
+ */
+struct MXTensor {
+ MXTensor() : data(nullptr) {}
+
+ MXTensor(void *data, const std::vector , MXDType dtype)
+ : data{data}, shape{shape}, dtype{dtype} {}
+
+ /*!
+ * \brief helper function to cast data pointer
+ */
+ template
+ data_type* getData() {
+return reinterpret_cast(data);
+ }
+
+ void *data; // not owned
+ std::vector shape;
+ MXDType dtype;
+};
+
+/*!
+ * \brief resource malloc function to allocate memory inside fcompute function
+ */
+typedef void* (*xpu_malloc_t)(void*, int);
+
+/*!
+ * \brief Class to provide resource APIs to FCompute
+ */
+class OpResource {
+ public:
+ OpResource(xpu_malloc_t xm, void* _xm) : xpu_malloc(xm), _xpu_malloc(_xm) {}
+
+ /*!
+ * \brief allocate memory controlled by MXNet
+ */
+ void* alloc(int size) {
+return xpu_malloc(_xpu_malloc, size);
+ }
+ private:
+ xpu_malloc_t xpu_malloc;
+ void* _xpu_malloc;
+};
+
+/*!
+ * \brief StatefulOp wrapper class to pass to backend OpState
+ */
+class CustomStatefulOpWrapper {
+ public:
+ CustomStatefulOpWrapper(void* inst) : instance(inst) {}
+ void* get_instance() { return instance; }
+ private:
+ void* instance;
+};
+
+/*!
+ * \brief An prototype interface class for library author creating stateful op
+ */
+class CustomStatefulOp {
+ public:
+ virtual void Forward() = 0;
Review comment:
Shouldnt we make the constructor virtual and = 0 too? Or is there somewhere
we need to be able to construct instances of the super class?
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
With regards,
Apache Git Services
[GitHub] [incubator-mxnet] samskalicky commented on a change in pull request #15921: [WIP] dynamic custom operator support
samskalicky commented on a change in pull request #15921: [WIP] dynamic custom
operator support
URL: https://github.com/apache/incubator-mxnet/pull/15921#discussion_r321955086
##
File path: include/mxnet/lib_api.h
##
@@ -18,33 +18,627 @@
*/
/*!
- * Copyright (c) 2015 by Contributors
+ * Copyright (c) 2019 by Contributors
* \file lib_api.h
* \brief APIs to interact with libraries
+ * This API specifies function prototypes to
+ * register custom ops for library authors
*/
+
#ifndef MXNET_LIB_API_H_
#define MXNET_LIB_API_H_
+#include
+#include
+#include
+#include
+
+#define MX_LIBRARY_VERSION 1
+
/*!
- * \brief Following are the APIs implemented in the external library
+ * \brief External Tensor data types
+ */
+enum MXDType {
+ kFloat32 = 0,
+ kFloat64 = 1,
+ kFloat16 = 2,
+ kUint8 = 3,
+ kInt32 = 4,
+ kInt8 = 5,
+ kInt64 = 6,
+};
+
+enum MXReturnValue {
+ MX_FAIL = 0,
+ MX_SUCCESS = 1,
+};
+
+/*!
+ * \brief External Tensor data structure
+ */
+struct MXTensor {
+ MXTensor() : data(nullptr) {}
+
+ MXTensor(void *data, const std::vector , MXDType dtype)
+ : data{data}, shape{shape}, dtype{dtype} {}
+
+ /*!
+ * \brief helper function to cast data pointer
+ */
+ template
+ data_type* getData() {
+return reinterpret_cast(data);
+ }
+
+ void *data; // not owned
+ std::vector shape;
+ MXDType dtype;
+};
+
+/*!
+ * \brief resource malloc function to allocate memory inside fcompute function
+ */
+typedef void* (*xpu_malloc_t)(void*, int);
+
+/*!
+ * \brief Class to provide resource APIs to FCompute
+ */
+class OpResource {
+ public:
+ OpResource(xpu_malloc_t xm, void* _xm) : xpu_malloc(xm), _xpu_malloc(_xm) {}
+
+ /*!
+ * \brief allocate memory controlled by MXNet
+ */
+ void* alloc(int size) {
+return xpu_malloc(_xpu_malloc, size);
+ }
+ private:
+ xpu_malloc_t xpu_malloc;
+ void* _xpu_malloc;
+};
+
+/*!
+ * \brief StatefulOp wrapper class to pass to backend OpState
+ */
+class CustomStatefulOpWrapper {
+ public:
+ CustomStatefulOpWrapper(void* inst) : instance(inst) {}
+ void* get_instance() { return instance; }
+ private:
+ void* instance;
+};
+
+/*!
+ * \brief An prototype interface class for library author creating stateful op
+ */
+class CustomStatefulOp {
+ public:
+ virtual void Forward() = 0;
+ virtual ~CustomStatefulOp() = 0;
+};
+
+/*!
+ * Custom Operator function templates
+ */
+typedef MXReturnValue (*fcomp_t)(std::map,
+ std::vector, std::vector,
+ OpResource res);
+typedef MXReturnValue (*parseAttrs_t)(std::map,
+ int*, int*);
+typedef MXReturnValue (*inferType_t)(std::map,
+ std::vector&, std::vector&);
+typedef MXReturnValue (*inferShape_t)(std::map,
+ std::vector>&,
+ std::vector>&);
+typedef MXReturnValue (*mutateInputs_t)(std::map,
+ std::vector&);
+typedef MXReturnValue (*createOpState_t)(std::map,
+ CustomStatefulOp**);
+typedef MXReturnValue (*fstateful_t)(CustomStatefulOp*, std::vector,
+ std::vector);
+
+/*!
+ * \brief Class to hold custom operator registration
+ */
+class CustomOp {
+ public:
+ explicit CustomOp(const char* op_name) : name(op_name), fcompute(nullptr),
+fgradient(nullptr), parse_attrs(nullptr), infer_type(nullptr),
infer_shape(nullptr),
+mutate_inputs(nullptr), create_op_state(nullptr), fstateful(nullptr) {}
+ ~CustomOp() {}
+ CustomOp& setForward(fcomp_t fcomp) {
+fcompute = fcomp;
+return *this;
+ }
+ CustomOp& setGradient(fcomp_t fcomp) {
+fgradient = fcomp;
+return *this;
+ }
+ CustomOp& setParseAttrs(parseAttrs_t func) {
+parse_attrs = func;
+return *this;
+ }
+ CustomOp& setInferType(inferType_t func) {
+infer_type = func;
+return *this;
+ }
+ CustomOp& setInferShape(inferShape_t func) {
+infer_shape = func;
+return *this;
+ }
+ CustomOp& setMutateInputs(mutateInputs_t func) {
+mutate_inputs = func;
+return *this;
+ }
+ CustomOp& setCreateOpState(createOpState_t func) {
+create_op_state = func;
+return *this;
+ }
+ CustomOp& setForwardStateful(fstateful_t func) {
+fstateful = func;
+return *this;
+ }
Review comment:
We need to add some error checking in lib_api.h to prevent users from
registering the wrong combo of functions. Currently we have 2 scenarios:
Basic Op Registration requires:
- Parse Attrs
- Infer Type
- Infer Shape
- Forward
__optional__
- Gradient --> Lets rename to Backward (to match with Forward)
- Mutate inputs
Stateful Op Registration requires:
- Parse Attrs
- Infer Type
- Infer Shape
- Create Op State
- Forward Stateful
__optional__
- Gradient --> Lets rename to Backward (to match
[GitHub] [incubator-mxnet] samskalicky commented on a change in pull request #15921: [WIP] dynamic custom operator support
samskalicky commented on a change in pull request #15921: [WIP] dynamic custom
operator support
URL: https://github.com/apache/incubator-mxnet/pull/15921#discussion_r321954968
##
File path: example/lib_ops/subgraph_lib.cc
##
@@ -0,0 +1,125 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements. See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership. The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License. You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing,
+ * software distributed under the License is distributed on an
+ * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+ * KIND, either express or implied. See the License for the
+ * specific language governing permissions and limitations
+ * under the License.
+ */
+
+/*!
+ * Copyright (c) 2019 by Contributors
+ * \file subgraph_lib.cc
+ * \brief subgraph operator implementation
+ * library file
+ */
+
+#include
+#include "lib_api.h"
+
+MXReturnValue parseAttrs(std::map attrs,
+ int* num_in, int* num_out) {
+ *num_in = 2;
+ *num_out = 1;
+ return MX_SUCCESS;
+}
+
+MXReturnValue inferType(std::map attrs,
std::vector ,
+ std::vector ) {
+ outtypes[0] = intypes[0];
+ return MX_SUCCESS;
+}
+
+MXReturnValue inferShape(std::map attrs,
std::vector> ,
+ std::vector> ) {
+ unsigned n = inshapes[0][0];
+ unsigned k = inshapes[0][1];
+ unsigned kk = inshapes[1][0];
+ unsigned m = inshapes[1][1];
+
+ std::cout << "inshapes[0][0]=" << n << " inshapes[0][1]=" << k << std::endl;
+ std::cout << "inshapes[1][0]=" << kk << " inshapes[1][1]=" << m <<
std::endl;
+
+ if (k != kk)
+return MX_FAIL;
+
+ outshapes[0].push_back(n);
+ outshapes[0].push_back(m);
+ return MX_SUCCESS;
+}
+
+MXReturnValue mutateInputs(std::map attrs,
+ std::vector _indices) {
+ input_indices.push_back(1);
+ std::cout << "the 1st input is marked as mutate input by library author" <<
std::endl;
+ return MX_SUCCESS;
+}
+
+class MyStatefulOp : public CustomStatefulOp {
+ public:
+ MyStatefulOp(std::string sym, int count) : subgraph_sym(sym), count(count) {}
+
+ void Forward() {
+count++;
+ }
+
+ int State() {
+return count;
+ }
+
+ ~MyStatefulOp() {}
+
+ private:
+ std::string subgraph_sym;
+ int count;
+};
+
+MXReturnValue createOpState(std::map attrs,
+CustomStatefulOp** op_inst) {
+ *op_inst = new MyStatefulOp("json", 0);
+ std::cout << "create op state successful" << std::endl;
+ return MX_SUCCESS;
+}
+
+MXReturnValue forwardStateful(CustomStatefulOp* op_inst,
+ std::vector inputs,
+ std::vector outputs) {
+ MyStatefulOp* my_op_inst = static_cast(op_inst);
+ if (my_op_inst == nullptr) {
+std::cout << "stateful op loading failed" << std::endl;
+return MX_FAIL;
+ }
+
+ my_op_inst->Forward();
Review comment:
Need to pass inputs/outputs to forward function
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
With regards,
Apache Git Services
[GitHub] [incubator-mxnet] samskalicky commented on a change in pull request #15921: [WIP] dynamic custom operator support
samskalicky commented on a change in pull request #15921: [WIP] dynamic custom
operator support
URL: https://github.com/apache/incubator-mxnet/pull/15921#discussion_r321954930
##
File path: include/mxnet/lib_api.h
##
@@ -18,33 +18,627 @@
*/
/*!
- * Copyright (c) 2015 by Contributors
+ * Copyright (c) 2019 by Contributors
* \file lib_api.h
* \brief APIs to interact with libraries
+ * This API specifies function prototypes to
+ * register custom ops for library authors
*/
+
#ifndef MXNET_LIB_API_H_
#define MXNET_LIB_API_H_
+#include
+#include
+#include
+#include
+
+#define MX_LIBRARY_VERSION 1
+
/*!
- * \brief Following are the APIs implemented in the external library
+ * \brief External Tensor data types
+ */
+enum MXDType {
+ kFloat32 = 0,
+ kFloat64 = 1,
+ kFloat16 = 2,
+ kUint8 = 3,
+ kInt32 = 4,
+ kInt8 = 5,
+ kInt64 = 6,
+};
+
+enum MXReturnValue {
+ MX_FAIL = 0,
+ MX_SUCCESS = 1,
+};
+
+/*!
+ * \brief External Tensor data structure
+ */
+struct MXTensor {
+ MXTensor() : data(nullptr) {}
+
+ MXTensor(void *data, const std::vector , MXDType dtype)
+ : data{data}, shape{shape}, dtype{dtype} {}
+
+ /*!
+ * \brief helper function to cast data pointer
+ */
+ template
+ data_type* getData() {
+return reinterpret_cast(data);
+ }
+
+ void *data; // not owned
+ std::vector shape;
+ MXDType dtype;
+};
+
+/*!
+ * \brief resource malloc function to allocate memory inside fcompute function
+ */
+typedef void* (*xpu_malloc_t)(void*, int);
+
+/*!
+ * \brief Class to provide resource APIs to FCompute
+ */
+class OpResource {
+ public:
+ OpResource(xpu_malloc_t xm, void* _xm) : xpu_malloc(xm), _xpu_malloc(_xm) {}
+
+ /*!
+ * \brief allocate memory controlled by MXNet
+ */
+ void* alloc(int size) {
+return xpu_malloc(_xpu_malloc, size);
+ }
+ private:
+ xpu_malloc_t xpu_malloc;
+ void* _xpu_malloc;
+};
+
+/*!
+ * \brief StatefulOp wrapper class to pass to backend OpState
+ */
+class CustomStatefulOpWrapper {
+ public:
+ CustomStatefulOpWrapper(void* inst) : instance(inst) {}
+ void* get_instance() { return instance; }
+ private:
+ void* instance;
+};
+
+/*!
+ * \brief An prototype interface class for library author creating stateful op
+ */
+class CustomStatefulOp {
+ public:
+ virtual void Forward() = 0;
+ virtual ~CustomStatefulOp() = 0;
+};
+
+/*!
+ * Custom Operator function templates
+ */
+typedef MXReturnValue (*fcomp_t)(std::map,
+ std::vector, std::vector,
+ OpResource res);
+typedef MXReturnValue (*parseAttrs_t)(std::map,
+ int*, int*);
+typedef MXReturnValue (*inferType_t)(std::map,
+ std::vector&, std::vector&);
+typedef MXReturnValue (*inferShape_t)(std::map,
+ std::vector>&,
+ std::vector>&);
+typedef MXReturnValue (*mutateInputs_t)(std::map,
+ std::vector&);
+typedef MXReturnValue (*createOpState_t)(std::map,
+ CustomStatefulOp**);
+typedef MXReturnValue (*fstateful_t)(CustomStatefulOp*, std::vector,
+ std::vector);
+
+/*!
+ * \brief Class to hold custom operator registration
+ */
+class CustomOp {
+ public:
+ explicit CustomOp(const char* op_name) : name(op_name), fcompute(nullptr),
+fgradient(nullptr), parse_attrs(nullptr), infer_type(nullptr),
infer_shape(nullptr),
+mutate_inputs(nullptr), create_op_state(nullptr), fstateful(nullptr) {}
+ ~CustomOp() {}
+ CustomOp& setForward(fcomp_t fcomp) {
+fcompute = fcomp;
+return *this;
+ }
+ CustomOp& setGradient(fcomp_t fcomp) {
+fgradient = fcomp;
+return *this;
+ }
+ CustomOp& setParseAttrs(parseAttrs_t func) {
+parse_attrs = func;
+return *this;
+ }
+ CustomOp& setInferType(inferType_t func) {
+infer_type = func;
+return *this;
+ }
+ CustomOp& setInferShape(inferShape_t func) {
+infer_shape = func;
+return *this;
+ }
+ CustomOp& setMutateInputs(mutateInputs_t func) {
+mutate_inputs = func;
+return *this;
+ }
+ CustomOp& setCreateOpState(createOpState_t func) {
+create_op_state = func;
+return *this;
+ }
+ CustomOp& setForwardStateful(fstateful_t func) {
+fstateful = func;
+return *this;
+ }
+ /*! \brief operator name */
+ const char* name;
+ /*! \brief operator functions */
+ fcomp_t fcompute;
+ fcomp_t fgradient;
+ parseAttrs_t parse_attrs;
+ inferType_t infer_type;
+ inferShape_t infer_shape;
+ mutateInputs_t mutate_inputs;
+ createOpState_t create_op_state;
+ fstateful_t fstateful;
+};
+
+/*!
+ * \brief Registry class to registers things (ops, properties)
+ * Singleton class
+ */
+template
+class Registry {
+ public:
+ /*!
+ * \brief get singleton pointer to class
+ * \returns pointer to class
+ */
+ static Registry* get() {
+
[GitHub] [incubator-mxnet] eric-haibin-lin commented on issue #16060: [bug] mxnet.ndarray.sparse.norm fallback regression in 1.5.0 and master
eric-haibin-lin commented on issue #16060: [bug] mxnet.ndarray.sparse.norm
fallback regression in 1.5.0 and master
URL:
https://github.com/apache/incubator-mxnet/issues/16060#issuecomment-529069438
It looks like the regression happens around April 16th
```
➜ mxnet git:(take) ✗ pip install mxnet==1.5.0b20190417
Requirement already satisfied: mxnet==1.5.0b20190417 in
/Users/haibilin/miniconda3/lib/python3.7/site-packages (1.5.0b20190417)
Requirement already satisfied: numpy<1.15.0,>=1.8.2 in
/Users/haibilin/miniconda3/lib/python3.7/site-packages (from
mxnet==1.5.0b20190417) (1.14.6)
Requirement already satisfied: requests>=2.20.0 in
/Users/haibilin/miniconda3/lib/python3.7/site-packages (from
mxnet==1.5.0b20190417) (2.22.0)
Requirement already satisfied: graphviz<0.9.0,>=0.8.1 in
/Users/haibilin/miniconda3/lib/python3.7/site-packages (from
mxnet==1.5.0b20190417) (0.8.4)
Requirement already satisfied: idna<2.9,>=2.5 in
/Users/haibilin/miniconda3/lib/python3.7/site-packages (from
requests>=2.20.0->mxnet==1.5.0b20190417) (2.8)
Requirement already satisfied: certifi>=2017.4.17 in
/Users/haibilin/miniconda3/lib/python3.7/site-packages (from
requests>=2.20.0->mxnet==1.5.0b20190417) (2019.6.16)
Requirement already satisfied: chardet<3.1.0,>=3.0.2 in
/Users/haibilin/miniconda3/lib/python3.7/site-packages (from
requests>=2.20.0->mxnet==1.5.0b20190417) (3.0.4)
Requirement already satisfied: urllib3!=1.25.0,!=1.25.1,<1.26,>=1.21.1 in
/Users/haibilin/miniconda3/lib/python3.7/site-packages (from
requests>=2.20.0->mxnet==1.5.0b20190417) (1.24.2)
➜ mxnet git:(take) ✗ python test.py
[20:54:47] src/operator/contrib/../tensor/../../common/utils.h:450:
Storage type fallback detected:
operator = norm
input storage types = [row_sparse, ]
output storage types = [default, ]
params = {}
context.dev_mask = cpu
The operator with default storage type will be dispatched for execution.
You're seeing this warning message because the operator above is unable to
process the given ndarrays with specified storage types, context and parameter.
Temporary dense ndarrays are generated in order to execute the operator. This
does not affect the correctness of the programme. You can set environment
variable MXNET_STORAGE_FALLBACK_LOG_VERBOSE to 0 to suppress this warning.
[2.]
➜ mxnet git:(take) ✗ pip install mxnet==1.5.0b20190416
Collecting mxnet==1.5.0b20190416
Using cached
https://files.pythonhosted.org/packages/48/41/99ca13c3173c3631a024ace26e36baedf7d0810c0ac465f22cc2f0af2796/mxnet-1.5.0b20190416-cp37-cp37m-macosx_10_11_x86_64.whl
Requirement already satisfied: numpy<1.15.0,>=1.8.2 in
/Users/haibilin/miniconda3/lib/python3.7/site-packages (from
mxnet==1.5.0b20190416) (1.14.6)
Requirement already satisfied: graphviz<0.9.0,>=0.8.1 in
/Users/haibilin/miniconda3/lib/python3.7/site-packages (from
mxnet==1.5.0b20190416) (0.8.4)
Requirement already satisfied: requests>=2.20.0 in
/Users/haibilin/miniconda3/lib/python3.7/site-packages (from
mxnet==1.5.0b20190416) (2.22.0)
Requirement already satisfied: certifi>=2017.4.17 in
/Users/haibilin/miniconda3/lib/python3.7/site-packages (from
requests>=2.20.0->mxnet==1.5.0b20190416) (2019.6.16)
Requirement already satisfied: idna<2.9,>=2.5 in
/Users/haibilin/miniconda3/lib/python3.7/site-packages (from
requests>=2.20.0->mxnet==1.5.0b20190416) (2.8)
Requirement already satisfied: chardet<3.1.0,>=3.0.2 in
/Users/haibilin/miniconda3/lib/python3.7/site-packages (from
requests>=2.20.0->mxnet==1.5.0b20190416) (3.0.4)
Requirement already satisfied: urllib3!=1.25.0,!=1.25.1,<1.26,>=1.21.1 in
/Users/haibilin/miniconda3/lib/python3.7/site-packages (from
requests>=2.20.0->mxnet==1.5.0b20190416) (1.24.2)
Installing collected packages: mxnet
Found existing installation: mxnet 1.5.0b20190417
Uninstalling mxnet-1.5.0b20190417:
Successfully uninstalled mxnet-1.5.0b20190417
Successfully installed mxnet-1.5.0b20190416
➜ mxnet git:(take) ✗ python test.py
[2.]
```
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
With regards,
Apache Git Services
[GitHub] [incubator-mxnet] yifeim commented on issue #16060: [bug] mxnet.ndarray.sparse.norm fallback regression in 1.5.0 and master
yifeim commented on issue #16060: [bug] mxnet.ndarray.sparse.norm fallback regression in 1.5.0 and master URL: https://github.com/apache/incubator-mxnet/issues/16060#issuecomment-529067609 @eric-haibin-lin This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] [incubator-mxnet] reminisce commented on issue #16097: [numpy] array ufunc and array function protocols
reminisce commented on issue #16097: [numpy] array ufunc and array function protocols URL: https://github.com/apache/incubator-mxnet/pull/16097#issuecomment-529063949 @szha This PR is ready for merge if there are no more comments. Thanks. This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[incubator-mxnet-site] branch asf-site updated: Bump the publish timestamp.
This is an automated email from the ASF dual-hosted git repository. marcoabreu pushed a commit to branch asf-site in repository https://gitbox.apache.org/repos/asf/incubator-mxnet-site.git The following commit(s) were added to refs/heads/asf-site by this push: new 948a3bc Bump the publish timestamp. 948a3bc is described below commit 948a3bc794da06e7d604596f1996c5ea802805cb Author: mxnet-ci AuthorDate: Sat Sep 7 01:29:05 2019 + Bump the publish timestamp. --- date.txt | 1 + 1 file changed, 1 insertion(+) diff --git a/date.txt b/date.txt new file mode 100644 index 000..2d4411c --- /dev/null +++ b/date.txt @@ -0,0 +1 @@ +Sat Sep 7 01:29:05 UTC 2019
[GitHub] [incubator-mxnet] anirudh2290 commented on issue #15148: Very Large CPU RAM Memory Consumption (>1GB)
anirudh2290 commented on issue #15148: Very Large CPU RAM Memory Consumption (>1GB) URL: https://github.com/apache/incubator-mxnet/issues/15148#issuecomment-529053611 We ( I and @karan6181 ) looked at a bunch of things to understand where the overhead is coming from. We looked at Resource Request and attach op resources pass. We looked at object pool in threaded engine, and at turning off OPENMP and MKLDNN and checking, but the overhead is still there. We looked at the overhead caused by different arrays in the module api. Overhead is not coming from any of these areas. We also check that increase in memory consumption happens at the bind stage and probably coming from somewhere in the graph executor. Next step is to check this in 1.2.1 to see if the increase happened in 1.3.1. This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] [incubator-mxnet] gigasquid commented on issue #16106: src/executor/graph_executor.cc:1847: Check failed: arg_names.size() == in_args_map.size() (2 vs. 1)
gigasquid commented on issue #16106: src/executor/graph_executor.cc:1847: Check failed: arg_names.size() == in_args_map.size() (2 vs. 1) URL: https://github.com/apache/incubator-mxnet/issues/16106#issuecomment-529051927 @adc17 looks reasonable to me. Thanks for helping on this This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] [incubator-mxnet] apeforest commented on a change in pull request #14535: [DOC] Updated install instructions for mac
apeforest commented on a change in pull request #14535: [DOC] Updated install instructions for mac URL: https://github.com/apache/incubator-mxnet/pull/14535#discussion_r321942658 ## File path: docs/install/osx_setup.md ## @@ -145,12 +155,16 @@ We have installed MXNet core library. Next, we will install MXNet interface pack To install the MXNet Python binding navigate to the root of the MXNet folder then run the following: ```bash -$ cd python -$ pip install -e . +virtualenv -p`which python3` mxnet_py3 +source mxnet_py3/bin/activate +pip install -e python ``` - +First we create a [virtualenv](https://docs.python-guide.org/dev/virtualenvs/#lower-level-virtualenv) to isolate this installation from our global environment. Review comment: This should not be required for installing mxnet. This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] [incubator-mxnet] marcoabreu commented on a change in pull request #14535: [DOC] Updated install instructions for mac
marcoabreu commented on a change in pull request #14535: [DOC] Updated install instructions for mac URL: https://github.com/apache/incubator-mxnet/pull/14535#discussion_r321942562 ## File path: docs/install/osx_setup.md ## @@ -97,17 +101,23 @@ Install the dependencies, required for MXNet, with the following commands: ### Build MXNet Shared Library After you have installed the dependencies, pull the MXNet source code from Git and build MXNet to produce an MXNet library called ```libmxnet.so```. You can clone the repository as described in the following code block, or you may try the [download links](download.md) for your desired MXNet version. -The file called ```osx.mk``` has the configuration required for building MXNet on OS X. First copy ```make/osx.mk``` into ```config.mk```, which is used by the ```make``` command: - ```bash -git clone --recursive https://github.com/apache/incubator-mxnet ~/mxnet -cd ~/mxnet -cp make/osx.mk ./config.mk -echo "USE_BLAS = openblas" >> ./config.mk -echo "ADD_CFLAGS += -I/usr/local/opt/openblas/include" >> ./config.mk -echo "ADD_LDFLAGS += -L/usr/local/opt/openblas/lib" >> ./config.mk -echo "ADD_LDFLAGS += -L/usr/local/lib/graphviz/" >> ./config.mk -make -j$(sysctl -n hw.ncpu) +git clone --recursive https://github.com/apache/incubator-mxnet ~/mxnet +cd ~/mxnet && pushd . +mkdir -p build && cd build +cmake \ +-DCMAKE_CXX_COMPILER_LAUNCHER=ccache \ +-DCMAKE_C_COMPILER_LAUNCHER=ccache \ +-DUSE_MKL_IF_AVAILABLE=OFF \ +-DUSE_MKLDNN=ON \ +-DUSE_CUDA=OFF \ +-DUSE_OPENMP=OFF \ +-DUSE_OPENCV=ON \ +-DUSE_SIGNAL_HANDLER=ON \ +-DCMAKE_BUILD_TYPE=Debug \ Review comment: Why are we making a debug build? Wouldn't a release build be a bit faster and improve the first impression? Or maybe release with symbols. This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] [incubator-mxnet] apeforest commented on a change in pull request #14535: [DOC] Updated install instructions for mac
apeforest commented on a change in pull request #14535: [DOC] Updated install instructions for mac URL: https://github.com/apache/incubator-mxnet/pull/14535#discussion_r321941620 ## File path: cmake/cmake_options_osx.yml ## @@ -0,0 +1,53 @@ +# Licensed to the Apache Software Foundation (ASF) under one Review comment: Instead of having this file, can we add a CMakeList_osx.txt instead? I think adding another layer of configurations on top of CMakeList.txt adds unnecessary effort to maintain. This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] [incubator-mxnet] marcoabreu commented on a change in pull request #14535: [DOC] Updated install instructions for mac
marcoabreu commented on a change in pull request #14535: [DOC] Updated install instructions for mac URL: https://github.com/apache/incubator-mxnet/pull/14535#discussion_r321942448 ## File path: docs/install/osx_setup.md ## @@ -97,17 +101,23 @@ Install the dependencies, required for MXNet, with the following commands: ### Build MXNet Shared Library After you have installed the dependencies, pull the MXNet source code from Git and build MXNet to produce an MXNet library called ```libmxnet.so```. You can clone the repository as described in the following code block, or you may try the [download links](download.md) for your desired MXNet version. -The file called ```osx.mk``` has the configuration required for building MXNet on OS X. First copy ```make/osx.mk``` into ```config.mk```, which is used by the ```make``` command: - ```bash -git clone --recursive https://github.com/apache/incubator-mxnet ~/mxnet -cd ~/mxnet -cp make/osx.mk ./config.mk -echo "USE_BLAS = openblas" >> ./config.mk -echo "ADD_CFLAGS += -I/usr/local/opt/openblas/include" >> ./config.mk -echo "ADD_LDFLAGS += -L/usr/local/opt/openblas/lib" >> ./config.mk -echo "ADD_LDFLAGS += -L/usr/local/lib/graphviz/" >> ./config.mk -make -j$(sysctl -n hw.ncpu) +git clone --recursive https://github.com/apache/incubator-mxnet ~/mxnet +cd ~/mxnet && pushd . +mkdir -p build && cd build +cmake \ +-DCMAKE_CXX_COMPILER_LAUNCHER=ccache \ +-DCMAKE_C_COMPILER_LAUNCHER=ccache \ +-DUSE_MKL_IF_AVAILABLE=OFF \ +-DUSE_MKLDNN=ON \ Review comment: Usually users shouldn't be touching cmakelists.txt This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] [incubator-mxnet] apeforest commented on a change in pull request #14535: [DOC] Updated install instructions for mac
apeforest commented on a change in pull request #14535: [DOC] Updated install instructions for mac URL: https://github.com/apache/incubator-mxnet/pull/14535#discussion_r321941620 ## File path: cmake/cmake_options_osx.yml ## @@ -0,0 +1,53 @@ +# Licensed to the Apache Software Foundation (ASF) under one Review comment: Instead of having this file, can we add a CMakeList_osx.txt instead? I think adding another layer on top of CMakeList.txt adds unnecessary effort to maintain. This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] [incubator-mxnet] apeforest commented on a change in pull request #14535: [DOC] Updated install instructions for mac
apeforest commented on a change in pull request #14535: [DOC] Updated install instructions for mac URL: https://github.com/apache/incubator-mxnet/pull/14535#discussion_r321941819 ## File path: cmake/cmake_options.yml ## @@ -46,7 +46,8 @@ USE_SIGNAL_HANDLER: "ON" # Print stack traces on segfaults. USE_TENSORRT: "OFF" # Enable infeference optimization with TensorRT. USE_ASAN: "OFF" # Enable Clang/GCC ASAN sanitizers. ENABLE_TESTCOVERAGE: "OFF" # Enable compilation with test coverage metric output -CMAKE_BUILD_TYPE: "Debug" +USE_INT64_TENSOR_SIZE: "OFF" # Use int64_t to represent the total number of elements in a tensor +CMAKE_BUILD_TYPE: "Debug" # Debug | Release | RelWithDebInfo | MinSizeRel Review comment: The danger of this default option is that users are not aware they are building a Debug instead of Release. Why not just let users do: ``` cmake -DCMAKE_BUILD_TYPE=Debug ``` This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] [incubator-mxnet] apeforest commented on a change in pull request #14535: [DOC] Updated install instructions for mac
apeforest commented on a change in pull request #14535: [DOC] Updated install instructions for mac URL: https://github.com/apache/incubator-mxnet/pull/14535#discussion_r321941620 ## File path: cmake/cmake_options_osx.yml ## @@ -0,0 +1,53 @@ +# Licensed to the Apache Software Foundation (ASF) under one Review comment: Instead of having this file, can we add a CMakeList_osx.txt instead? I think adding another layer on top of CMakeList.txt is unnecessary and difficult to maintain. This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] [incubator-mxnet] marcoabreu commented on a change in pull request #14535: [DOC] Updated install instructions for mac
marcoabreu commented on a change in pull request #14535: [DOC] Updated install instructions for mac URL: https://github.com/apache/incubator-mxnet/pull/14535#discussion_r321942344 ## File path: docs/install/osx_setup.md ## @@ -89,25 +92,31 @@ Install the dependencies, required for MXNet, with the following commands: # Get pip easy_install pip # For visualization of network graphs - pip install graphviz + pip3 install graphviz # Jupyter notebook - pip install jupyter + pip3 install jupyter ``` ### Build MXNet Shared Library After you have installed the dependencies, pull the MXNet source code from Git and build MXNet to produce an MXNet library called ```libmxnet.so```. You can clone the repository as described in the following code block, or you may try the download links for your desired MXNet version. -The file called ```osx.mk``` has the configuration required for building MXNet on OS X. First copy ```make/osx.mk``` into ```config.mk```, which is used by the ```make``` command: - ```bash -git clone --recursive https://github.com/apache/incubator-mxnet ~/mxnet -cd ~/mxnet -cp make/osx.mk ./config.mk -echo "USE_BLAS = openblas" >> ./config.mk -echo "ADD_CFLAGS += -I/usr/local/opt/openblas/include" >> ./config.mk -echo "ADD_LDFLAGS += -L/usr/local/opt/openblas/lib" >> ./config.mk -echo "ADD_LDFLAGS += -L/usr/local/lib/graphviz/" >> ./config.mk -make -j$(sysctl -n hw.ncpu) +git clone --recursive https://github.com/apache/incubator-mxnet ~/mxnet +cd ~/mxnet && pushd . +mkdir build && cd build +cmake \ +-DCMAKE_CXX_COMPILER_LAUNCHER=ccache \ +-DCMAKE_C_COMPILER_LAUNCHER=ccache \ +-DUSE_MKL_IF_AVAILABLE=OFF \ Review comment: @pengzhao-intel are these the optimal compile flags? Especially with regard to omp and mkl? Also, why are we so verbose here? Are our defaults not good enough? This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] [incubator-mxnet] apeforest commented on a change in pull request #14535: [DOC] Updated install instructions for mac
apeforest commented on a change in pull request #14535: [DOC] Updated install instructions for mac URL: https://github.com/apache/incubator-mxnet/pull/14535#discussion_r321942026 ## File path: docs/install/osx_setup.md ## @@ -97,17 +101,23 @@ Install the dependencies, required for MXNet, with the following commands: ### Build MXNet Shared Library After you have installed the dependencies, pull the MXNet source code from Git and build MXNet to produce an MXNet library called ```libmxnet.so```. You can clone the repository as described in the following code block, or you may try the [download links](download.md) for your desired MXNet version. -The file called ```osx.mk``` has the configuration required for building MXNet on OS X. First copy ```make/osx.mk``` into ```config.mk```, which is used by the ```make``` command: - ```bash -git clone --recursive https://github.com/apache/incubator-mxnet ~/mxnet -cd ~/mxnet -cp make/osx.mk ./config.mk -echo "USE_BLAS = openblas" >> ./config.mk -echo "ADD_CFLAGS += -I/usr/local/opt/openblas/include" >> ./config.mk -echo "ADD_LDFLAGS += -L/usr/local/opt/openblas/lib" >> ./config.mk -echo "ADD_LDFLAGS += -L/usr/local/lib/graphviz/" >> ./config.mk -make -j$(sysctl -n hw.ncpu) +git clone --recursive https://github.com/apache/incubator-mxnet ~/mxnet +cd ~/mxnet && pushd . +mkdir -p build && cd build +cmake \ +-DCMAKE_CXX_COMPILER_LAUNCHER=ccache \ +-DCMAKE_C_COMPILER_LAUNCHER=ccache \ +-DUSE_MKL_IF_AVAILABLE=OFF \ +-DUSE_MKLDNN=ON \ Review comment: I feel all these flags are unnecessary. We already have a CMakeList.txt and isn't that the place for user to switch on/off these flags? This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] [incubator-mxnet] apeforest commented on a change in pull request #14535: [DOC] Updated install instructions for mac
apeforest commented on a change in pull request #14535: [DOC] Updated install instructions for mac URL: https://github.com/apache/incubator-mxnet/pull/14535#discussion_r321941914 ## File path: docs/install/osx_setup.md ## @@ -97,17 +101,23 @@ Install the dependencies, required for MXNet, with the following commands: ### Build MXNet Shared Library After you have installed the dependencies, pull the MXNet source code from Git and build MXNet to produce an MXNet library called ```libmxnet.so```. You can clone the repository as described in the following code block, or you may try the [download links](download.md) for your desired MXNet version. -The file called ```osx.mk``` has the configuration required for building MXNet on OS X. First copy ```make/osx.mk``` into ```config.mk```, which is used by the ```make``` command: - ```bash -git clone --recursive https://github.com/apache/incubator-mxnet ~/mxnet -cd ~/mxnet -cp make/osx.mk ./config.mk -echo "USE_BLAS = openblas" >> ./config.mk -echo "ADD_CFLAGS += -I/usr/local/opt/openblas/include" >> ./config.mk -echo "ADD_LDFLAGS += -L/usr/local/opt/openblas/lib" >> ./config.mk -echo "ADD_LDFLAGS += -L/usr/local/lib/graphviz/" >> ./config.mk -make -j$(sysctl -n hw.ncpu) +git clone --recursive https://github.com/apache/incubator-mxnet ~/mxnet +cd ~/mxnet && pushd . +mkdir -p build && cd build +cmake \ +-DCMAKE_CXX_COMPILER_LAUNCHER=ccache \ +-DCMAKE_C_COMPILER_LAUNCHER=ccache \ +-DUSE_MKL_IF_AVAILABLE=OFF \ +-DUSE_MKLDNN=ON \ +-DUSE_CUDA=OFF \ +-DUSE_OPENMP=OFF \ +-DUSE_OPENCV=ON \ +-DUSE_SIGNAL_HANDLER=ON \ +-DCMAKE_BUILD_TYPE=Debug \ +-GNinja .. +ninja +popd Review comment: Why need `pushd .` and `popd`? This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] [incubator-mxnet] apeforest commented on a change in pull request #14535: [DOC] Updated install instructions for mac
apeforest commented on a change in pull request #14535: [DOC] Updated install instructions for mac URL: https://github.com/apache/incubator-mxnet/pull/14535#discussion_r321941819 ## File path: cmake/cmake_options.yml ## @@ -46,7 +46,8 @@ USE_SIGNAL_HANDLER: "ON" # Print stack traces on segfaults. USE_TENSORRT: "OFF" # Enable infeference optimization with TensorRT. USE_ASAN: "OFF" # Enable Clang/GCC ASAN sanitizers. ENABLE_TESTCOVERAGE: "OFF" # Enable compilation with test coverage metric output -CMAKE_BUILD_TYPE: "Debug" +USE_INT64_TENSOR_SIZE: "OFF" # Use int64_t to represent the total number of elements in a tensor +CMAKE_BUILD_TYPE: "Debug" # Debug | Release | RelWithDebInfo | MinSizeRel Review comment: The danger of this default option is that users are not aware they are building a Debug instead of Release. This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] [incubator-mxnet] apeforest commented on a change in pull request #14535: [DOC] Updated install instructions for mac
apeforest commented on a change in pull request #14535: [DOC] Updated install instructions for mac URL: https://github.com/apache/incubator-mxnet/pull/14535#discussion_r321941620 ## File path: cmake/cmake_options_osx.yml ## @@ -0,0 +1,53 @@ +# Licensed to the Apache Software Foundation (ASF) under one Review comment: Instead of having this file, can we add a CMakeList_osx.txt instead? I think this adding another layer on top of CMakeList.txt is unnecessary. This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] [incubator-mxnet] marcoabreu commented on a change in pull request #14535: [DOC] Updated install instructions for mac
marcoabreu commented on a change in pull request #14535: [DOC] Updated install instructions for mac URL: https://github.com/apache/incubator-mxnet/pull/14535#discussion_r321941326 ## File path: docs/install/osx_setup.md ## @@ -145,12 +155,16 @@ We have installed MXNet core library. Next, we will install MXNet interface pack To install the MXNet Python binding navigate to the root of the MXNet folder then run the following: ```bash -$ cd python -$ pip install -e . +virtualenv -p`which python3` mxnet_py3 Review comment: While I generally agree that virtualenv is a good practice, it adds another dependency. These guides are for the very basic user to get something running. I think this PR overcomplicates a lot of things. Adding the cmake compilation? Fine! Everything else? Please leave it as it is. This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] [incubator-mxnet] marcoabreu commented on a change in pull request #14535: [DOC] Updated install instructions for mac
marcoabreu commented on a change in pull request #14535: [DOC] Updated install instructions for mac URL: https://github.com/apache/incubator-mxnet/pull/14535#discussion_r321941068 ## File path: docs/install/osx_setup.md ## @@ -100,14 +104,21 @@ After you have installed the dependencies, pull the MXNet source code from Git a The file called ```osx.mk``` has the configuration required for building MXNet on OS X. First copy ```make/osx.mk``` into ```config.mk```, which is used by the ```make``` command: ```bash -git clone --recursive https://github.com/apache/incubator-mxnet ~/mxnet -cd ~/mxnet -cp make/osx.mk ./config.mk -echo "USE_BLAS = openblas" >> ./config.mk -echo "ADD_CFLAGS += -I/usr/local/opt/openblas/include" >> ./config.mk -echo "ADD_LDFLAGS += -L/usr/local/opt/openblas/lib" >> ./config.mk -echo "ADD_LDFLAGS += -L/usr/local/lib/graphviz/" >> ./config.mk -make -j$(sysctl -n hw.ncpu) +git clone --recursive https://github.com/apache/incubator-mxnet ~/mxnet +cd ~/mxnet +mkdir build && cd build +cmake \ +-DCMAKE_CXX_COMPILER_LAUNCHER=ccache \ Review comment: These should be bare minimum steps. Requiring ccache is an advanced thing and I'd prefer to not have it in the basic instructions. This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] [incubator-mxnet] apeforest commented on a change in pull request #14535: [DOC] Updated install instructions for mac
apeforest commented on a change in pull request #14535: [DOC] Updated install instructions for mac URL: https://github.com/apache/incubator-mxnet/pull/14535#discussion_r321939887 ## File path: docs/install/osx_setup.md ## @@ -77,6 +77,10 @@ Install the dependencies, required for MXNet, with the following commands: ```bash brew update +brew install python3 Review comment: Should this be python instead of python3? Also it seems the community has not reached an agreement on retiring python2 sypport yet. This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] [incubator-mxnet] karan6181 commented on issue #15148: Very Large CPU RAM Memory Consumption (>1GB)
karan6181 commented on issue #15148: Very Large CPU RAM Memory Consumption
(>1GB)
URL:
https://github.com/apache/incubator-mxnet/issues/15148#issuecomment-529044425
- I see roughly 2 GB of CPU memory usage when I ran the user script and also
with the below minimum reproducible script using Gluon.
- ```python
import mxnet as mx
import gluoncv
from time import sleep
net = gluoncv.model_zoo.get_model('cifar_resnet20_v1', pretrained=True,
ctx=mx.gpu())
net.hybridize()
#print(net.summary(mx.nd.ones((1, 3,28,28), ctx=mx.gpu(
sleep(20)
```
- I tried using different MXNet version but the CPU memory usage keeps on
increasing. Not Much difference in memory consumption if using with/without MKL.
- ```bash
# P3.16xLarge
# DLAMI V24
MXNet-cu92: 1.3.1: 2.15 GB
MXNet-cu92: 1.4.1: 2.3 GB
MXNet-cu92: 1.5.0: 2.55 GB
MXNet-cu92: 1.6.0(1.6.0b20190906): 2.9 GB
```
- If I run the same above script on CPU context, then the memory usage is
approx. 100 MB.
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
With regards,
Apache Git Services
[GitHub] [incubator-mxnet] apeforest merged pull request #15746: [MXNET-978] Higher Order Gradient Support `clip`, `dropout`.
apeforest merged pull request #15746: [MXNET-978] Higher Order Gradient Support `clip`, `dropout`. URL: https://github.com/apache/incubator-mxnet/pull/15746 This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[incubator-mxnet] branch master updated (c928392 -> 24f0a10)
This is an automated email from the ASF dual-hosted git repository. apeforest pushed a change to branch master in repository https://gitbox.apache.org/repos/asf/incubator-mxnet.git. from c928392 [MXNET-978] Higher Order Gradient Support `sqrt`, `cbrt`. (#15474) add 24f0a10 [MXNET-978] Higher Order Gradient Support `clip`, `dropout`. (#15746) No new revisions were added by this update. Summary of changes: src/operator/nn/dropout.cc | 4 +++- src/operator/tensor/matrix_op.cc| 3 ++- tests/python/unittest/test_higher_order_grad.py | 30 + 3 files changed, 35 insertions(+), 2 deletions(-)
[incubator-mxnet] branch master updated (6de6848 -> c928392)
This is an automated email from the ASF dual-hosted git repository. apeforest pushed a change to branch master in repository https://gitbox.apache.org/repos/asf/incubator-mxnet.git. from 6de6848 Not to search for coverage files when none exist (#16107) add c928392 [MXNET-978] Higher Order Gradient Support `sqrt`, `cbrt`. (#15474) No new revisions were added by this update. Summary of changes: src/operator/tensor/elemwise_unary_op_pow.cc| 71 - tests/python/unittest/test_higher_order_grad.py | 40 ++ 2 files changed, 109 insertions(+), 2 deletions(-)
[GitHub] [incubator-mxnet] apeforest merged pull request #15474: [MXNET-978] Higher Order Gradient Support `sqrt`, `cbrt`.
apeforest merged pull request #15474: [MXNET-978] Higher Order Gradient Support `sqrt`, `cbrt`. URL: https://github.com/apache/incubator-mxnet/pull/15474 This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] [incubator-mxnet] adc17 commented on issue #16106: src/executor/graph_executor.cc:1847: Check failed: arg_names.size() == in_args_map.size() (2 vs. 1)
adc17 commented on issue #16106: src/executor/graph_executor.cc:1847: Check failed: arg_names.size() == in_args_map.size() (2 vs. 1) URL: https://github.com/apache/incubator-mxnet/issues/16106#issuecomment-529039677 @ZhennanQin The Clojure test cases that failed are `test-maximum` and `test-minimum`: https://github.com/apache/incubator-mxnet/blob/6de684825130c28dfa75f2a707aeaed64a4340e5/contrib/clojure-package/test/org/apache/clojure_mxnet/operator_test.clj#L409-L441 Changing the duplicate `"data"` strings to `"data1"` and `"data2"` resolves the failures . @gigasquid I think the correct fix is to update the tests + add an error message for this scenario? I will submit a PR if you agree. This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] [incubator-mxnet] zhreshold commented on issue #16114: improve dataloader signals and messages
zhreshold commented on issue #16114: improve dataloader signals and messages URL: https://github.com/apache/incubator-mxnet/pull/16114#issuecomment-529036509 @szha, @eric-haibin-lin @sxjscience for review This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] [incubator-mxnet] zhreshold opened a new pull request #16114: improve dataloader signals and messages
zhreshold opened a new pull request #16114: improve dataloader signals and messages URL: https://github.com/apache/incubator-mxnet/pull/16114 ## Description ## Improve dataloader use experience. With this PR, dataloaders are - More responsive to terminate with Ctrl + C - Show helpful messages if exceptions raises in multiple workers. DataLoaders used to produce unreadable messages. - Less likely to hang forever, given timeout is added to fetch logic ## Checklist ## ### Essentials ### Please feel free to remove inapplicable items for your PR. - [ ] The PR title starts with [MXNET-$JIRA_ID], where $JIRA_ID refers to the relevant [JIRA issue](https://issues.apache.org/jira/projects/MXNET/issues) created (except PRs with tiny changes) - [x] Changes are complete (i.e. I finished coding on this PR) - [x] All changes have test coverage: - Unit tests are added for small changes to verify correctness (e.g. adding a new operator) - Nightly tests are added for complicated/long-running ones (e.g. changing distributed kvstore) - Build tests will be added for build configuration changes (e.g. adding a new build option with NCCL) - [ ] Code is well-documented: - For user-facing API changes, API doc string has been updated. - For new C++ functions in header files, their functionalities and arguments are documented. - For new examples, README.md is added to explain the what the example does, the source of the dataset, expected performance on test set and reference to the original paper if applicable - Check the API doc at http://mxnet-ci-doc.s3-accelerate.dualstack.amazonaws.com/PR-$PR_ID/$BUILD_ID/index.html - [ ] To the my best knowledge, examples are either not affected by this change, or have been fixed to be compatible with this change ### Changes ### - [ ] Feature1, tests, (and when applicable, API doc) - [ ] Feature2, tests, (and when applicable, API doc) ## Comments ## - If this change is a backward incompatible change, why must this change be made. - Interesting edge cases to note here This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] [incubator-mxnet] mxnet-label-bot commented on issue #16113: [Flaky] test_mkldnn.test_activation
mxnet-label-bot commented on issue #16113: [Flaky] test_mkldnn.test_activation URL: https://github.com/apache/incubator-mxnet/issues/16113#issuecomment-529035181 Hey, this is the MXNet Label Bot. Thank you for submitting the issue! I will try and suggest some labels so that the appropriate MXNet community members can help resolve it. Here are my recommended label(s): Test, Flaky This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] [incubator-mxnet] reminisce opened a new issue #16113: [Flaky] test_mkldnn.test_activation
reminisce opened a new issue #16113: [Flaky] test_mkldnn.test_activation
URL: https://github.com/apache/incubator-mxnet/issues/16113
```
==
FAIL: test_mkldnn.test_activation
--
Traceback (most recent call last):
File "/usr/local/lib/python3.5/dist-packages/nose/case.py", line 198, in
runTest
self.test(*self.arg)
File "/work/mxnet/tests/python/mkl/../unittest/common.py", line 177, in
test_new
orig_test(*args, **kwargs)
File "/work/mxnet/tests/python/mkl/test_mkldnn.py", line 350, in
test_activation
check_activation_training(stype)
File "/work/mxnet/tests/python/mkl/test_mkldnn.py", line 346, in
check_activation_training
check_numeric_gradient(test, in_location, numeric_eps=1e-5, rtol=0.16,
atol=1e-4)
File "/work/mxnet/python/mxnet/test_utils.py", line 1015, in
check_numeric_gradient
("NUMERICAL_%s"%name, "BACKWARD_%s"%name))
File "/work/mxnet/python/mxnet/test_utils.py", line 533, in
assert_almost_equal
raise AssertionError(msg)
AssertionError:
Items are not equal:
Error 1.609589 exceeds tolerance rtol=0.16, atol=0.000100. Location of
maximum error:(0, 1, 0, 0), a=0.523432, b=0.705208
NUMERICAL_data: array(0.1385808 , 0.],
[0.923872 , 0.58710575]],
...
BACKWARD_data: array(0.13833651, 0.],
[0.9269223 , 0.58610183]],
...
```
@PatricZhao Any idea? Thanks.
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
With regards,
Apache Git Services
[GitHub] [incubator-mxnet] marcoabreu commented on a change in pull request #15808: Add option to choose between OMP implementations
marcoabreu commented on a change in pull request #15808: Add option to choose
between OMP implementations
URL: https://github.com/apache/incubator-mxnet/pull/15808#discussion_r321929537
##
File path: CMakeLists.txt
##
@@ -432,14 +432,13 @@ endif()
# ---[ OpenMP
if(USE_OPENMP)
- find_package(OpenMP REQUIRED)
# This should build on Windows, but there's some problem and I don't have a
Windows box, so
# could a Windows user please fix?
- if(EXISTS ${CMAKE_CURRENT_SOURCE_DIR}/3rdparty/openmp/CMakeLists.txt
- AND SYSTEM_ARCHITECTURE STREQUAL "x86_64"
- AND NOT MSVC
- AND NOT CMAKE_CROSSCOMPILING)
-
+ if(USE_OPENMP STREQUAL "BUNDLED" AND EXISTS
${CMAKE_CURRENT_SOURCE_DIR}/3rdparty/openmp/CMakeLists.txt
Review comment:
The first check and the others should be separate to throw appropriate error
messages
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
With regards,
Apache Git Services
[GitHub] [incubator-mxnet] marcoabreu commented on a change in pull request #15808: Add option to choose between OMP implementations
marcoabreu commented on a change in pull request #15808: Add option to choose
between OMP implementations
URL: https://github.com/apache/incubator-mxnet/pull/15808#discussion_r321929582
##
File path: CMakeLists.txt
##
@@ -452,14 +451,18 @@ if(USE_OPENMP)
set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} ${OpenMP_C_FLAGS}")
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} ${OpenMP_CXX_FLAGS}")
add_definitions(-DMXNET_USE_OPENMP=1)
- else()
+ elseif(USE_OPENMP STREQUAL "PLATFORM" OR USE_OPENMP STREQUAL "ON")
Review comment:
What's the difference between on and platform
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
With regards,
Apache Git Services
[GitHub] [incubator-mxnet] access2rohit opened a new pull request #16112: [DO NOT MERGE]Revert custom profiler
access2rohit opened a new pull request #16112: [DO NOT MERGE]Revert custom profiler URL: https://github.com/apache/incubator-mxnet/pull/16112 ## Description ## (Brief description on what this PR is about) ## Checklist ## ### Essentials ### Please feel free to remove inapplicable items for your PR. - [ ] The PR title starts with [MXNET-$JIRA_ID], where $JIRA_ID refers to the relevant [JIRA issue](https://issues.apache.org/jira/projects/MXNET/issues) created (except PRs with tiny changes) - [ ] Changes are complete (i.e. I finished coding on this PR) - [ ] All changes have test coverage: - Unit tests are added for small changes to verify correctness (e.g. adding a new operator) - Nightly tests are added for complicated/long-running ones (e.g. changing distributed kvstore) - Build tests will be added for build configuration changes (e.g. adding a new build option with NCCL) - [ ] Code is well-documented: - For user-facing API changes, API doc string has been updated. - For new C++ functions in header files, their functionalities and arguments are documented. - For new examples, README.md is added to explain the what the example does, the source of the dataset, expected performance on test set and reference to the original paper if applicable - Check the API doc at http://mxnet-ci-doc.s3-accelerate.dualstack.amazonaws.com/PR-$PR_ID/$BUILD_ID/index.html - [ ] To the my best knowledge, examples are either not affected by this change, or have been fixed to be compatible with this change ### Changes ### - [ ] Feature1, tests, (and when applicable, API doc) - [ ] Feature2, tests, (and when applicable, API doc) ## Comments ## - If this change is a backward incompatible change, why must this change be made. - Interesting edge cases to note here This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] [incubator-mxnet] apeforest commented on issue #15808: Add option to choose between OMP implementations
apeforest commented on issue #15808: Add option to choose between OMP implementations URL: https://github.com/apache/incubator-mxnet/pull/15808#issuecomment-529030769 > They are for CI, a change in dmlc is needed first: > > [dmlc/dmlc-core#558](https://github.com/dmlc/dmlc-core/pull/558) Why is it needed for this PR? This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] [incubator-mxnet] apeforest commented on a change in pull request #15808: Add option to choose between OMP implementations
apeforest commented on a change in pull request #15808: Add option to choose
between OMP implementations
URL: https://github.com/apache/incubator-mxnet/pull/15808#discussion_r321926506
##
File path: CMakeLists.txt
##
@@ -452,14 +451,18 @@ if(USE_OPENMP)
set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} ${OpenMP_C_FLAGS}")
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} ${OpenMP_CXX_FLAGS}")
add_definitions(-DMXNET_USE_OPENMP=1)
- else()
+ elseif(USE_OPENMP STREQUAL "PLATFORM" OR USE_OPENMP STREQUAL "ON")
+find_package(OpenMP REQUIRED)
+message("Using platform provided OpenMP")
if(OPENMP_FOUND)
set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} ${OpenMP_C_FLAGS}")
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} ${OpenMP_CXX_FLAGS}")
set(CMAKE_SHARED_LINKER_FLAGS "${CMAKE_SHARED_LINKER_FLAGS}
${OpenMP_EXE_LINKER_FLAGS}")
set(CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS}
${OpenMP_EXE_LINKER_FLAGS}")
add_definitions(-DMXNET_USE_OPENMP=1)
endif()
+ else()
+message(FATAL_ERROR "USE_OPENMP takes values [PLATFORM, BUNDLED, OFF]")
Review comment:
"USE_OPENMP takes values [PLATFORM, BUNDLED, OFF, ON]")?
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
With regards,
Apache Git Services
[GitHub] [incubator-mxnet] apeforest commented on a change in pull request #15808: Add option to choose between OMP implementations
apeforest commented on a change in pull request #15808: Add option to choose between OMP implementations URL: https://github.com/apache/incubator-mxnet/pull/15808#discussion_r321926437 ## File path: CMakeLists.txt ## @@ -23,7 +23,7 @@ mxnet_option(USE_CUDA "Build with CUDA support" ON) mxnet_option(USE_OLDCMAKECUDA "Build with old cmake cuda" OFF) mxnet_option(USE_NCCL "Use NVidia NCCL with CUDA" OFF) mxnet_option(USE_OPENCV "Build with OpenCV support" ON) -mxnet_option(USE_OPENMP "Build with Openmp support" ON) +mxnet_option(USE_OPENMP "Build with Openmp support" ON) # OFF | ON | PLATFORM | BUNDLED Review comment: What's the difference between ON and the rest? Can you add some comments here? This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] [incubator-mxnet] sxjscience commented on issue #16111: Incompatible data type for numpy ops?
sxjscience commented on issue #16111: Incompatible data type for numpy ops? URL: https://github.com/apache/incubator-mxnet/issues/16111#issuecomment-529019794 Link this with https://github.com/apache/incubator-mxnet/issues/16048 This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] [incubator-mxnet] sxjscience commented on issue #16111: Incompatible data type for numpy ops?
sxjscience commented on issue #16111: Incompatible data type for numpy ops? URL: https://github.com/apache/incubator-mxnet/issues/16111#issuecomment-529017466 @reminisce I think it's reasonable. This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[incubator-mxnet-site] branch asf-site updated: Bump the publish timestamp.
This is an automated email from the ASF dual-hosted git repository. marcoabreu pushed a commit to branch asf-site in repository https://gitbox.apache.org/repos/asf/incubator-mxnet-site.git The following commit(s) were added to refs/heads/asf-site by this push: new a108875 Bump the publish timestamp. a108875 is described below commit a10887577808f46452bbc0fdea3a3cefdf51f31b Author: mxnet-ci AuthorDate: Fri Sep 6 21:09:57 2019 + Bump the publish timestamp. --- date.txt | 1 + 1 file changed, 1 insertion(+) diff --git a/date.txt b/date.txt new file mode 100644 index 000..35ddab6 --- /dev/null +++ b/date.txt @@ -0,0 +1 @@ +Fri Sep 6 21:09:57 UTC 2019
[GitHub] [incubator-mxnet] apeforest commented on issue #16023: Revert "Refactor LibraryInitializer so it's thread safe. Fixes random sporadical concurrency crashes."
apeforest commented on issue #16023: Revert "Refactor LibraryInitializer so it's thread safe. Fixes random sporadical concurrency crashes." URL: https://github.com/apache/incubator-mxnet/pull/16023#issuecomment-529011468 https://github.com/apache/incubator-mxnet/pull/16040 already reverted the unintentional change in the original PR. So I am closing this one. This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] [incubator-mxnet] apeforest closed pull request #16023: Revert "Refactor LibraryInitializer so it's thread safe. Fixes random sporadical concurrency crashes."
apeforest closed pull request #16023: Revert "Refactor LibraryInitializer so it's thread safe. Fixes random sporadical concurrency crashes." URL: https://github.com/apache/incubator-mxnet/pull/16023 This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] [incubator-mxnet] reminisce commented on issue #16111: Incompatible data type for numpy ops?
reminisce commented on issue #16111: Incompatible data type for numpy ops? URL: https://github.com/apache/incubator-mxnet/issues/16111#issuecomment-529010437 Yeah, we are already aware of this and has been working on this. It requires us to re-write the kernel of those ops and that would become much less tedious and efforts after we use TVM op module developed by @yzhliu . One catch here is that we may not be able to provide backward compute function for int input types as the backward pass shares the same infer type function as the forward. So for those ops taking integer inputs, we give error message that the gradient calculation is not supported and ask users to use floats as inputs. What do you guys think? This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] [incubator-mxnet] cjolivier01 removed a comment on issue #15167: Pointwise fusion for GPU
cjolivier01 removed a comment on issue #15167: Pointwise fusion for GPU URL: https://github.com/apache/incubator-mxnet/pull/15167#issuecomment-529008979 I am removing my block on this PR and letting in this "auto-abuse" in since I am in a similar industry to NVidia and don't wish to make it look like I am blocking a performance improvement for NVidia GPU's. I hope that the auto-abuse will be fixed before merging, however. This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] [incubator-mxnet] cjolivier01 commented on issue #15167: Pointwise fusion for GPU
cjolivier01 commented on issue #15167: Pointwise fusion for GPU URL: https://github.com/apache/incubator-mxnet/pull/15167#issuecomment-529008979 I am removing my block on this PR and letting in this "auto-abuse" in since I am in a similar industry to NVidia and don't wish to make it look like I am blocking a performance improvement for NVidia GPU's. I hope that the auto-abuse will be fixed before merging, however. This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] [incubator-mxnet] larroy commented on issue #15808: Add option to choose between OMP implementations
larroy commented on issue #15808: Add option to choose between OMP implementations URL: https://github.com/apache/incubator-mxnet/pull/15808#issuecomment-529008630 @lebeg please approve then This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] [incubator-mxnet] eric-haibin-lin commented on issue #16111: Incompatible data type for numpy ops?
eric-haibin-lin commented on issue #16111: Incompatible data type for numpy ops? URL: https://github.com/apache/incubator-mxnet/issues/16111#issuecomment-529007400 @reminisce @sxjscience This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] [incubator-mxnet] mxnet-label-bot commented on issue #16111: Incompatible data type for numpy ops?
mxnet-label-bot commented on issue #16111: Incompatible data type for numpy ops? URL: https://github.com/apache/incubator-mxnet/issues/16111#issuecomment-529006579 Hey, this is the MXNet Label Bot. Thank you for submitting the issue! I will try and suggest some labels so that the appropriate MXNet community members can help resolve it. Here are my recommended label(s): Feature This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] [incubator-mxnet] eric-haibin-lin opened a new issue #16111: Incompatible data type for numpy ops?
eric-haibin-lin opened a new issue #16111: Incompatible data type for numpy ops? URL: https://github.com/apache/incubator-mxnet/issues/16111 ``` >>> mx.np.log(mx.np.array([1], dtype='int')) array([0], dtype=int64) >>> np.log(np.array([10])) array([2.30258509]) ``` Many math operators in numpy (such as `log`) do not return integer result. However, `mx.np` does. Is this expected? This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] [incubator-mxnet] apeforest commented on issue #15811: [MXNET-891] Support tuple of scales in upsample operator
apeforest commented on issue #15811: [MXNET-891] Support tuple of scales in upsample operator URL: https://github.com/apache/incubator-mxnet/pull/15811#issuecomment-529003895 Ping @huangzhiyuan for review. This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[incubator-mxnet] branch master updated (259f6bb -> 6de6848)
This is an automated email from the ASF dual-hosted git repository. yuxihu pushed a change to branch master in repository https://gitbox.apache.org/repos/asf/incubator-mxnet.git. from 259f6bb Revert accidental change to CMakelists (#16040) add 6de6848 Not to search for coverage files when none exist (#16107) No new revisions were added by this update. Summary of changes: tests/nightly/JenkinsfileForBinaries | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
[GitHub] [incubator-mxnet] yuxihu merged pull request #16107: Skip coverage files find for nightly tests
yuxihu merged pull request #16107: Skip coverage files find for nightly tests URL: https://github.com/apache/incubator-mxnet/pull/16107 This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] [incubator-mxnet] apeforest merged pull request #16040: Revert accidental change to CMakelists
apeforest merged pull request #16040: Revert accidental change to CMakelists URL: https://github.com/apache/incubator-mxnet/pull/16040 This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[incubator-mxnet] branch master updated (255dff0 -> 259f6bb)
This is an automated email from the ASF dual-hosted git repository. apeforest pushed a change to branch master in repository https://gitbox.apache.org/repos/asf/incubator-mxnet.git. from 255dff0 [MXNET-978] Higher Order Gradient Support `arctan`, `arctanh`, `radians`. (#15531) add 259f6bb Revert accidental change to CMakelists (#16040) No new revisions were added by this update. Summary of changes: CMakeLists.txt | 12 +--- 1 file changed, 5 insertions(+), 7 deletions(-)
[incubator-mxnet-site] branch asf-site updated: Bump the publish timestamp.
This is an automated email from the ASF dual-hosted git repository. marcoabreu pushed a commit to branch asf-site in repository https://gitbox.apache.org/repos/asf/incubator-mxnet-site.git The following commit(s) were added to refs/heads/asf-site by this push: new c60e023 Bump the publish timestamp. c60e023 is described below commit c60e0234929ca90d90149810564592b81ec34cbf Author: mxnet-ci AuthorDate: Fri Sep 6 19:30:01 2019 + Bump the publish timestamp. --- date.txt | 1 + 1 file changed, 1 insertion(+) diff --git a/date.txt b/date.txt new file mode 100644 index 000..df98cef --- /dev/null +++ b/date.txt @@ -0,0 +1 @@ +Fri Sep 6 19:30:01 UTC 2019
[incubator-mxnet] branch master updated (d85a2d0 -> 255dff0)
This is an automated email from the ASF dual-hosted git repository. apeforest pushed a change to branch master in repository https://gitbox.apache.org/repos/asf/incubator-mxnet.git. from d85a2d0 [DOC] Fix doc for nn.Embedding, nn.Dense and nd.Embedding (#15869) add 255dff0 [MXNET-978] Higher Order Gradient Support `arctan`, `arctanh`, `radians`. (#15531) No new revisions were added by this update. Summary of changes: src/nnvm/node_op_util.h | 76 + src/operator/tensor/elemwise_unary_op_trig.cc | 63 +++- tests/python/unittest/test_higher_order_grad.py | 46 +++ 3 files changed, 182 insertions(+), 3 deletions(-) create mode 100644 src/nnvm/node_op_util.h
[GitHub] [incubator-mxnet] apeforest merged pull request #15531: [MXNET-978] Higher Order Gradient Support `arctan`, `arctanh`, `radians`.
apeforest merged pull request #15531: [MXNET-978] Higher Order Gradient Support `arctan`, `arctanh`, `radians`. URL: https://github.com/apache/incubator-mxnet/pull/15531 This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] [incubator-mxnet] apeforest edited a comment on issue #16104: Faster Transpose 2D
apeforest edited a comment on issue #16104: Faster Transpose 2D URL: https://github.com/apache/incubator-mxnet/pull/16104#issuecomment-528957235 > Here numpy transpose calls the same TransposeImpl function that I have already handled. Yes, please check why your code broke their unit test? This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] [incubator-mxnet] apeforest commented on issue #16104: Faster Transpose 2D
apeforest commented on issue #16104: Faster Transpose 2D URL: https://github.com/apache/incubator-mxnet/pull/16104#issuecomment-528957235 > Here numpy transpose calls the same TransposeImpl function that I have already handled. Yes, please check why your code breaks their unit test? This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] [incubator-mxnet] reminisce commented on issue #16038: Delete printing messages in unit tests
reminisce commented on issue #16038: Delete printing messages in unit tests URL: https://github.com/apache/incubator-mxnet/pull/16038#issuecomment-528956965 Cleaned up in another PR. This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] [incubator-mxnet] reminisce closed pull request #16038: Delete printing messages in unit tests
reminisce closed pull request #16038: Delete printing messages in unit tests URL: https://github.com/apache/incubator-mxnet/pull/16038 This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] [incubator-mxnet] apeforest commented on a change in pull request #16104: Faster Transpose 2D
apeforest commented on a change in pull request #16104: Faster Transpose 2D
URL: https://github.com/apache/incubator-mxnet/pull/16104#discussion_r321850068
##
File path: src/operator/tensor/matrix_op-inl.h
##
@@ -257,6 +257,29 @@ struct TransposeParam : public
dmlc::Parameter {
}
};
+
+template
+MSHADOW_XINLINE void Transpose2D(DType *in, DType *out, index_t shape_0,
index_t shape_1) {
+// ensure cache line hits and prevent cache miss for any configuration
+index_t blocksize = 32;
+index_t n = shape_0;
+index_t p = shape_1;
+
+for (index_t i = 0; i < n; i += blocksize) {
+ #pragma omp parallel for
Review comment:
Why OMP parallel for the innerloop only?
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
With regards,
Apache Git Services
[GitHub] [incubator-mxnet] apeforest commented on a change in pull request #16104: Faster Transpose 2D
apeforest commented on a change in pull request #16104: Faster Transpose 2D
URL: https://github.com/apache/incubator-mxnet/pull/16104#discussion_r321849931
##
File path: src/operator/tensor/matrix_op-inl.h
##
@@ -257,6 +257,29 @@ struct TransposeParam : public
dmlc::Parameter {
}
};
+
+template
+MSHADOW_XINLINE void Transpose2D(DType *in, DType *out, index_t shape_0,
index_t shape_1) {
+// ensure cache line hits and prevent cache miss for any configuration
+index_t blocksize = 32;
+index_t n = shape_0;
+index_t p = shape_1;
+
+for (index_t i = 0; i < n; i += blocksize) {
Review comment:
How is this blocksize decided? Is it platform dependent?
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
With regards,
Apache Git Services
[GitHub] [incubator-mxnet] HahTK edited a comment on issue #16108: Subgraph API creates duplicate inputs and outputs
HahTK edited a comment on issue #16108: Subgraph API creates duplicate inputs
and outputs
URL:
https://github.com/apache/incubator-mxnet/issues/16108#issuecomment-528951277
Two things to add :
1. A temporary fix has been tested for fixing the output duplication. (code
below)
This fix fits things into the current function signatures and had to redo
similar computation in 2 spots
2. A better fix would be to introduce a map for output_entries and
input_entries to allow 1-many and many-1 mappings between the subgraph and the
maingraph. Conceptually, similar with the fix in (1)
Code :
src/operator/subgraph/partition_graph.cc:CreateSubgraphNode()
```
nnvm::Symbol sym;
nnvm::NodeEntryEqual node_equal;
size_t idx = 0;
sym.outputs.resize(output_entries.size());
//only add unique output_entries to sym.outputs
//relies on output_entries being pre-sorted
for (size_t i = 0; i < output_entries.size(); ++i) {
if (0 == i) {
sym.outputs[idx] = *output_entries[i];
} else {
if (!node_equal(*output_entries[i-1], *output_entries[i])) {
idx++;
sym.outputs[idx] = *output_entries[i];
} //else skip over dupe entry
}
}
sym.outputs.resize(idx+1);
```
In src/operator/subgraph/subgraph_property.h : ConnectSubgraphOutputs()
```
nnvm::NodeEntryEqual node_equal;
nnvm::NodeEntry prevNodeEntry;
uint32_t idx = 0;
//increment NodeEntry index only if the output_entry is unique
for (size_t i = 0; i < output_entries->size(); ++i) {
if (0 != i ) {
if (!node_equal(prevNodeEntry, *output_entries->at(i))) {
idx++;
}
}
prevNodeEntry = *output_entries->at(i); //need a copy
*output_entries->at(i) = nnvm::NodeEntry{subgraph_node, idx, 0};
}
```
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
With regards,
Apache Git Services
[GitHub] [incubator-mxnet] HahTK edited a comment on issue #16108: Subgraph API creates duplicate inputs and outputs
HahTK edited a comment on issue #16108: Subgraph API creates duplicate inputs
and outputs
URL:
https://github.com/apache/incubator-mxnet/issues/16108#issuecomment-528951277
Two things to add :
1. A temporary fix has been tested for fixing the output duplication. (code
below)
This fix fits things into the current function signatures and had to redo
similar computation in 2 spots
2. A better fix would be to introduce a map for output_entries and
input_entries to allow 1-many and many-1 mappings between the subgraph and the
maingraph. Conceptually, similar with the fix in (1)
Code :
src/operator/subgraph/partition_graph.cc:CreateSubgraphNode()
```
nnvm::Symbol sym;
nnvm::NodeEntryEqual node_equal;
size_t idx = 0;
sym.outputs.resize(output_entries.size());
//only add unique output_entries to sym.outputs
//relies on output_entries being pre-sorted
for (size_t i = 0; i < output_entries.size(); ++i) {
if (0 == i) {
sym.outputs[idx] = *output_entries[i];
} else {
if (!node_equal(*output_entries[i-1], *output_entries[i])) {
idx++;
sym.outputs[idx] = *output_entries[i];
} //else skip over dupe entry
}
}
sym.outputs.resize(idx+1);
```
In src/operator/subgraph/subgraph_property.h : ConnectSubgraphOutputs()
```
nnvm::NodeEntryEqual node_equal;
nnvm::NodeEntry prevNodeEntry;
uint32_t idx = 0;
//increment NodeEntry index only if the output_entrie is unique
for (size_t i = 0; i < output_entries->size(); ++i) {
if (0 != i ) {
if (!node_equal(prevNodeEntry, *output_entries->at(i))) {
idx++;
}
}
prevNodeEntry = *output_entries->at(i); //need a copy
*output_entries->at(i) = nnvm::NodeEntry{subgraph_node, idx, 0};
}
```
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
With regards,
Apache Git Services
[GitHub] [incubator-mxnet] HahTK edited a comment on issue #16108: Subgraph API creates duplicate inputs and outputs
HahTK edited a comment on issue #16108: Subgraph API creates duplicate inputs
and outputs
URL:
https://github.com/apache/incubator-mxnet/issues/16108#issuecomment-528951277
Two things to add :
1. A temporary fix has been tested for fixing the output duplication. (code
below)
This fix fits things into the current function signatures and had to redo
similar computation in 2 spots
2. A better fix would be to introduce a map for output_entries and
input_entries to allow 1-many and many-1 mappings between the subgraph and the
maingraph. Conceptually, similar with the fix in (1)
Code :
src/operator/subgraph/partition_graph.cc:CreateSubgraphNode()
```
nnvm::Symbol sym;
nnvm::NodeEntryEqual node_equal;
size_t idx = 0;
sym.outputs.resize(output_entries.size());
//only add to output_entries to sym.outputs
//relies on output_entries being pre-sorted
for (size_t i = 0; i < output_entries.size(); ++i) {
if (0 == i) {
sym.outputs[idx] = *output_entries[i];
} else {
if (!node_equal(*output_entries[i-1], *output_entries[i])) {
idx++;
sym.outputs[idx] = *output_entries[i];
} //else skip over dupe entry
}
}
sym.outputs.resize(idx+1);
```
In src/operator/subgraph/subgraph_property.h : ConnectSubgraphOutputs()
```
nnvm::NodeEntryEqual node_equal;
nnvm::NodeEntry prevNodeEntry;
uint32_t idx = 0;
//increment NodeEntry index only if the output_entrie is unique
for (size_t i = 0; i < output_entries->size(); ++i) {
if (0 != i ) {
if (!node_equal(prevNodeEntry, *output_entries->at(i))) {
idx++;
}
}
prevNodeEntry = *output_entries->at(i); //need a copy
*output_entries->at(i) = nnvm::NodeEntry{subgraph_node, idx, 0};
}
```
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
With regards,
Apache Git Services
[GitHub] [incubator-mxnet] HahTK edited a comment on issue #16108: Subgraph API creates duplicate inputs and outputs
HahTK edited a comment on issue #16108: Subgraph API creates duplicate inputs
and outputs
URL:
https://github.com/apache/incubator-mxnet/issues/16108#issuecomment-528951277
Two things to add :
1. A temporary fix has been tested for fixing the output duplication. (code
below)
This fix fits things into the current function signatures and had to redo
similar computation in 2 spots
2. A better fix would be to introduce a map for output_entries and
input_entries to allow 1-many and many-1 mappings between the subgraph and the
maingraph. Conceptually, similar with the fix in (1)
Code :
src/operator/subgraph/partition_graph.cc:CreateSubgraphNode()
nnvm::Symbol sym;
nnvm::NodeEntryEqual node_equal;
size_t idx = 0;
sym.outputs.resize(output_entries.size());
//only add to output_entries to sym.outputs
//relies on output_entries being pre-sorted
for (size_t i = 0; i < output_entries.size(); ++i) {
if (0 == i) {
sym.outputs[idx] = *output_entries[i];
} else {
if (!node_equal(*output_entries[i-1], *output_entries[i])) {
idx++;
sym.outputs[idx] = *output_entries[i];
} //else skip over dupe entry
}
}
sym.outputs.resize(idx+1);
In src/operator/subgraph/subgraph_property.h : ConnectSubgraphOutputs()
nnvm::NodeEntryEqual node_equal;
nnvm::NodeEntry prevNodeEntry;
uint32_t idx = 0;
//increment NodeEntry index only if the output_entrie is unique
for (size_t i = 0; i < output_entries->size(); ++i) {
if (0 != i ) {
if (!node_equal(prevNodeEntry, *output_entries->at(i))) {
idx++;
}
}
prevNodeEntry = *output_entries->at(i); //need a copy
*output_entries->at(i) = nnvm::NodeEntry{subgraph_node, idx, 0};
}
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
With regards,
Apache Git Services
[GitHub] [incubator-mxnet] HahTK edited a comment on issue #16108: Subgraph API creates duplicate inputs and outputs
HahTK edited a comment on issue #16108: Subgraph API creates duplicate inputs
and outputs
URL:
https://github.com/apache/incubator-mxnet/issues/16108#issuecomment-528951277
Two things to add :
1. A temporary fix has been tested for fixing the output duplication. (code
below)
This fix fits things into the current function signatures and had to redo
similar computation in 2 spot
2. A better fix would be to introduce a map for output_entries and
input_entries to allow 1-many and many-1 mappings between the subgraph and the
maingraph. Conceptually, similar with the fix in (1)
Code :
src/operator/subgraph/partition_graph.cc:CreateSubgraphNode()
nnvm::Symbol sym;
nnvm::NodeEntryEqual node_equal;
size_t idx = 0;
sym.outputs.resize(output_entries.size());
//only add to output_entries to sym.outputs
//relies on output_entries being pre-sorted
for (size_t i = 0; i < output_entries.size(); ++i) {
if (0 == i) {
sym.outputs[idx] = *output_entries[i];
} else {
if (!node_equal(*output_entries[i-1], *output_entries[i])) {
idx++;
sym.outputs[idx] = *output_entries[i];
} //else skip over dupe entry
}
}
sym.outputs.resize(idx+1);
In src/operator/subgraph/subgraph_property.h : ConnectSubgraphOutputs()
nnvm::NodeEntryEqual node_equal;
nnvm::NodeEntry prevNodeEntry;
uint32_t idx = 0;
//increment NodeEntry index only if the output_entrie is unique
for (size_t i = 0; i < output_entries->size(); ++i) {
if (0 != i ) {
if (!node_equal(prevNodeEntry, *output_entries->at(i))) {
idx++;
}
}
prevNodeEntry = *output_entries->at(i); //need a copy
*output_entries->at(i) = nnvm::NodeEntry{subgraph_node, idx, 0};
}
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
With regards,
Apache Git Services
[GitHub] [incubator-mxnet] apeforest commented on a change in pull request #16104: Faster Transpose 2D
apeforest commented on a change in pull request #16104: Faster Transpose 2D
URL: https://github.com/apache/incubator-mxnet/pull/16104#discussion_r321845284
##
File path: src/operator/tensor/matrix_op-inl.h
##
@@ -257,6 +257,29 @@ struct TransposeParam : public
dmlc::Parameter {
}
};
+
+template
+MSHADOW_XINLINE void Transpose2D(DType *in, DType *out, index_t shape_0,
index_t shape_1) {
+// ensure cache line hits and prevent cache miss for any configuration
+index_t blocksize = 32;
+index_t n = shape_0;
+index_t p = shape_1;
+
+for (index_t i = 0; i < n; i += blocksize) {
+ #pragma omp parallel for
+for (index_t j = 0; j < p; j += blocksize) {
+// transpose the block
+#pragma unroll 4
Review comment:
Do we really have to unroll this mannually? I thought the compiler could
figure out the optimal unrolling factor.
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
With regards,
Apache Git Services
[GitHub] [incubator-mxnet] HahTK commented on issue #16108: Subgraph API creates duplicate inputs and outputs
HahTK commented on issue #16108: Subgraph API creates duplicate inputs and
outputs
URL:
https://github.com/apache/incubator-mxnet/issues/16108#issuecomment-528951277
Two things to add :
1. A temporary fix has been tested for fixing the output duplication. (code
below)
This fix fits things into the current function signatures and had to redo
similar computation in 2 spot
2. A better fix would be to introduce a map for output_entries and
input_entries to allow 1-many and many-1 mappings between the subgraph and the
maingraph. Conceptually, similar with the fix in #1.
Code :
src/operator/subgraph/partition_graph.cc:CreateSubgraphNode()
nnvm::Symbol sym;
nnvm::NodeEntryEqual node_equal;
size_t idx = 0;
sym.outputs.resize(output_entries.size());
//only add to output_entries to sym.outputs
//relies on output_entries being pre-sorted
for (size_t i = 0; i < output_entries.size(); ++i) {
if (0 == i) {
sym.outputs[idx] = *output_entries[i];
} else {
if (!node_equal(*output_entries[i-1], *output_entries[i])) {
idx++;
sym.outputs[idx] = *output_entries[i];
} //else skip over dupe entry
}
}
sym.outputs.resize(idx+1);
In src/operator/subgraph/subgraph_property.h : ConnectSubgraphOutputs()
nnvm::NodeEntryEqual node_equal;
nnvm::NodeEntry prevNodeEntry;
uint32_t idx = 0;
//increment NodeEntry index only if the output_entrie is unique
for (size_t i = 0; i < output_entries->size(); ++i) {
if (0 != i ) {
if (!node_equal(prevNodeEntry, *output_entries->at(i))) {
idx++;
}
}
prevNodeEntry = *output_entries->at(i); //need a copy
*output_entries->at(i) = nnvm::NodeEntry{subgraph_node, idx, 0};
}
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
With regards,
Apache Git Services
[GitHub] [incubator-mxnet] access2rohit commented on a change in pull request #16104: Faster Transpose 2D
access2rohit commented on a change in pull request #16104: Faster Transpose 2D
URL: https://github.com/apache/incubator-mxnet/pull/16104#discussion_r321840156
##
File path: src/operator/tensor/matrix_op-inl.h
##
@@ -257,6 +257,29 @@ struct TransposeParam : public
dmlc::Parameter {
}
};
+
+template
Review comment:
Can you add comments explaining input params
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
With regards,
Apache Git Services
[incubator-mxnet] branch master updated (e98dbe7 -> d85a2d0)
This is an automated email from the ASF dual-hosted git repository. haibin pushed a change to branch master in repository https://gitbox.apache.org/repos/asf/incubator-mxnet.git. from e98dbe7 Speed up group executor (#16069) add d85a2d0 [DOC] Fix doc for nn.Embedding, nn.Dense and nd.Embedding (#15869) No new revisions were added by this update. Summary of changes: python/mxnet/gluon/nn/basic_layers.py | 17 ++--- src/operator/tensor/indexing_op.cc| 5 +++-- 2 files changed, 13 insertions(+), 9 deletions(-)
[GitHub] [incubator-mxnet] eric-haibin-lin merged pull request #15869: [DOC] Fix doc for nn.Embedding, nn.Dense and nd.Embedding
eric-haibin-lin merged pull request #15869: [DOC] Fix doc for nn.Embedding, nn.Dense and nd.Embedding URL: https://github.com/apache/incubator-mxnet/pull/15869 This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] [incubator-mxnet] hzfan commented on issue #16100: Infra for tvm op runtime dispatch
hzfan commented on issue #16100: Infra for tvm op runtime dispatch URL: https://github.com/apache/incubator-mxnet/pull/16100#issuecomment-528916168 > Just for curious. Based on my knowledge, tvm op kernel is pre-compiled and then linked together with MXNet. How can it be configured according to the runtime input shapes? Yes, kernels are pre-compiled. At compile time, several different schedules (kernels) for a single op are defined and compiled. Then at runtime, with the runtime input shape, the most suitable kernel is chosen. It's true that the kernel is pre-compiled, but we have multiple available kernels for one single op, so we can choose the most efficient one based on the runtime input shape. This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] [incubator-mxnet] ChaiBapchya commented on issue #16104: Faster Transpose 2D
ChaiBapchya commented on issue #16104: Faster Transpose 2D URL: https://github.com/apache/incubator-mxnet/pull/16104#issuecomment-528900889 Yes. @wuxun-zhang This week, we found out that benchmark shouldn't be run with DEBUG=ON. Turns out that after disabling the debug, results might be different. So I'll rerun the benchmarks for all configurations with DEBUG=OFF or Release mode (basically) and update the tracking issue. As of this table, you're right. MKL doesn't see performance difference for 2D input. This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[incubator-mxnet] branch master updated: Speed up group executor (#16069)
This is an automated email from the ASF dual-hosted git repository.
marcoabreu pushed a commit to branch master
in repository https://gitbox.apache.org/repos/asf/incubator-mxnet.git
The following commit(s) were added to refs/heads/master by this push:
new e98dbe7 Speed up group executor (#16069)
e98dbe7 is described below
commit e98dbe776201424cff93f87e67ff3bddb87d45e5
Author: Doron Singer <48903991+doronsin...@users.noreply.github.com>
AuthorDate: Fri Sep 6 17:53:04 2019 +0300
Speed up group executor (#16069)
* Speed up group executor
Current implementation is O(n^2), this implementation is O(n)
* Speed up group executor
Current implementation is O(n^2), this implementation is O(n)
* CI
---
python/mxnet/module/executor_group.py | 6 +++---
1 file changed, 3 insertions(+), 3 deletions(-)
diff --git a/python/mxnet/module/executor_group.py
b/python/mxnet/module/executor_group.py
index 637acce..d47665d 100755
--- a/python/mxnet/module/executor_group.py
+++ b/python/mxnet/module/executor_group.py
@@ -273,9 +273,9 @@ class DataParallelExecutorGroup(object):
self.data_layouts = None
self.label_layouts = None
self.output_names = self.symbol.list_outputs()
-self.output_layouts =
[DataDesc.get_batch_axis(self.symbol[name].attr('__layout__'))
- for name in self.output_names]
-self.num_outputs = len(self.symbol.list_outputs())
+self.num_outputs = len(self.output_names)
+self.output_layouts =
[DataDesc.get_batch_axis(self.symbol[index].attr('__layout__'))
+ for index in range(self.num_outputs)]
self.bind_exec(data_shapes, label_shapes, shared_group)
[GitHub] [incubator-mxnet] marcoabreu merged pull request #16069: Speed up group executor
marcoabreu merged pull request #16069: Speed up group executor URL: https://github.com/apache/incubator-mxnet/pull/16069 This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[incubator-mxnet] 01/01: Merge remote-tracking branch 'origin/master' into mkldnn-v1.0
This is an automated email from the ASF dual-hosted git repository.
taolv pushed a commit to branch mkldnn-v1.0
in repository https://gitbox.apache.org/repos/asf/incubator-mxnet.git
commit 5c5a619afaf69d607fbd254a585584b87f5b1dd8
Merge: 03b734b 7f57e8e
Author: Tao Lv
AuthorDate: Sat Sep 7 07:40:24 2019 +0800
Merge remote-tracking branch 'origin/master' into mkldnn-v1.0
Conflicts:
ci/jenkins/Jenkins_steps.groovy
3rdparty/mshadow/mshadow/cuda/tensor_gpu-inl.cuh | 26 +-
3rdparty/mshadow/mshadow/extension/slice.h |4 +-
3rdparty/mshadow/mshadow/tensor.h |4 +-
3rdparty/mshadow/mshadow/tensor_cpu-inl.h | 11 +-
3rdparty/mshadow/mshadow/tensor_gpu-inl.h |4 +-
3rdparty/ps-lite |2 +-
CMakeLists.txt |4 +-
CONTRIBUTORS.md|1 +
KEYS | 58 +
Makefile | 39 +-
README.md |1 +
benchmark/opperf/utils/benchmark_utils.py |2 +-
benchmark/opperf/utils/common_utils.py | 17 +-
benchmark/opperf/utils/profiler_utils.py | 29 +-
cd/Jenkinsfile_cd_pipeline | 62 +
cd/Jenkinsfile_release_job | 99 +
cd/Jenkinsfile_utils.groovy| 101 +
cd/README.md | 181 ++
cd/mxnet_lib/mxnet_lib_pipeline.groovy | 168 ++
cd/mxnet_lib/static/Jenkins_pipeline.groovy| 59 +
cd/utils/artifact_repository.md| 105 +
cd/utils/artifact_repository.py| 619 +
cd/utils/requirements.txt |2 +
cd/utils/test_artifact_repository.py | 530 +
ci/Jenkinsfile_utils.groovy|7 +-
ci/build_windows.py| 165 +-
ci/docker/Dockerfile.build.ubuntu_gpu_cu101|1 +
ci/docker/install/ubuntu_python.sh |2 +-
ci/docker/runtime_functions.sh | 107 +-
ci/jenkins/Jenkins_steps.groovy| 139 +-
ci/jenkins/Jenkinsfile_clang |4 +-
.../{Jenkinsfile_clang => Jenkinsfile_tools} | 14 +-
cmake/cmake_options.yml|1 -
contrib/clojure-package/README.md |8 +-
.../examples/profiler/test/core_test.clj |3 +-
.../profiler/test/profile-matmul-20iter.json.ref | 271 ---
contrib/clojure-package/integration-tests.sh |2 +-
contrib/clojure-package/project.clj|4 +-
docs/api/python/contrib/onnx.md|2 +-
docs/conf.py |2 +-
docs/cpp_docs/Doxyfile | 2370 +++
.../integration-tests.sh => docs/cpp_docs/Makefile | 20 +-
docs/install/build_from_source.md |2 +-
docs/python_docs/README.md | 24 +
docs/python_docs/_static/apache_incubator_logo.png | Bin 0 -> 16552 bytes
docs/python_docs/_static/google_analytics.js | 26 +
docs/python_docs/_static/minima-social-icons.svg | 33 +
docs/python_docs/_static/mxnet-icon.png| Bin 0 -> 2741 bytes
docs/python_docs/_static/mxnet.css | 199 ++
docs/python_docs/_static/mxnet_logo.png| Bin 0 -> 22390 bytes
.../python_docs/environment.yml| 36 +-
docs/python_docs/python/.gitignore | 20 +
docs/python_docs/python/Makefile | 57 +
docs/python_docs/python/Makefile_sphinx| 216 ++
docs/python_docs/python/README.md | 130 ++
docs/python_docs/python/api/advanced/index.rst | 74 +
.../python/api/advanced/mxnet.engine.rst | 34 +
.../python/api/advanced/mxnet.executor.rst | 34 +
.../python/api/advanced/mxnet.executor_manager.rst | 38 +
.../python/api/advanced/mxnet.kvstore_server.rst | 36 +
docs/python_docs/python/api/advanced/mxnet.rtc.rst | 36 +
.../python/api/advanced/mxnet.test_utils.rst | 91 +
.../python_docs/python/api/advanced/mxnet.util.rst | 31 +
.../python_docs/python/api/gluon-related/index.rst | 111 +
.../python/api/gluon-related/mxnet.autograd.rst| 38 +
.../python/api/gluon-related/mxnet.context.rst | 33 +
.../python/api/gluon-related/mxnet.image.rst | 99 +
.../python/api/gluon-related/mxnet.initializer.rst | 58 +
.../python/api/gluon-related/mxnet.io.rst | 52 +
.../api/gluon-related/mxnet.kvstore.KVStore.rst| 61 +
.../api/gluon-related/mxnet.kvstore.create.rst | 23 +
.../python/api/gluon-related/mxnet.kvstore.rst | 27 +
.../api/gluon-related/mxnet.lr_scheduler.rst | 31 +
[incubator-mxnet] branch mkldnn-v1.0 updated (03b734b -> 5c5a619)
This is an automated email from the ASF dual-hosted git repository.
taolv pushed a change to branch mkldnn-v1.0
in repository https://gitbox.apache.org/repos/asf/incubator-mxnet.git.
from 03b734b Merge remote-tracking branch 'origin' into mkldnn-v1.0
add 72c180c Correct ONNX documentation (#15914)
add acc074f Add AMP Conversion support for BucketingModule (#15528)
add ab60214 Add Median,p50,p99 to python profiler (#15953)
add 79ed678 [Numpy] random.randint() implemented (#15956)
add 0e71fbd Added tests to verify Large Vector Support for initial set of
ops (#15943)
add 8df9469 Refines NDArray indexing and adds numpy ndarray indexing
[READY FOR REVIEW] (#15942)
add b2c0cbc Windows cmake flags cleanup (#16013)
add 9b906a5 Improve diagnose.py to display environment variables (#15715)
add 649429d Disable flaky test in test_amp_conversion (#16031)
add 3f7b6ee Improve quantization flow (#15961)
add 2d86c70 Port ops from np branch (#16018)
add 196d1f4 [MXNET-1399] multiclass-mcc metric enhancements (#14874)
add b7cca01 [MXNET-895] ONNX import/export: TopK (#13627)
add 61f3dbc numpy-compatible cumsum upstream (#15924)
add 5d0d335 Update README.md (#16035)
add 36455b2 Add RROIAlign (#16017)
add 35d943e Updates git_init Jenkins utility function to support checking
out a particular commit id
add 1597498 Adds artifact repository scripts
add 87207c5 Adds CD pipeline framework
add b73b8d4 Adds static libmxnet release pipeline
add 0ed97f1 Updates CD pipeline
add 5fe1516 Adds documentation
add 1196c15 Updates kvstore functions to use pushd and popd
add 23a7a58 Throws exceptions instead o magic numbers
add e539370 Updates artifact repository cli to use --libtype instead of
--static or --dynamic
add fff8c82 Clarifies ci_utils and cd_utils origin remark
add 0570892 Adds clarifying note on why ubuntu 14.04 is being used for
compilation
add 2b12c59 Removes MXNET_SHA
add a8c0fe8 Removes set_release_job_name
add 5cb26fd Adds license headers
add 98cdf30 Updates artifact repository to expect licenses
add 3027296 Moves ci/cd to cd directory
add f6d0fc2 Takes downstream job name from environment
add 8241c52 Updates order of parameters
add 749492f Updates job type parameter to dropdown
add 759e76e Adds libmxnet feature extraction code comments
add 36ac85a Removes ccache setup from static build
add 65928b1 NumPy-compatible infrastructure on Gluon (#16024)
add 9173dad [MKLDNN] fix uint8 batch norm memory misuse (#16034)
add 47f8ceb Disable test coverage of C++ codebase on CI (#15981)
add 54d27cb [OP] Support range as advanced index for ndarrays (#16047)
add d80510a Added more tests for Large Indices (#15960)
add aab4ded Numpy compatible max min (#16046)
add 6997691 [Dev] update ps-lite dependency (#15936)
add 36bab1c Fix flaky clojure profile test (#16058)
add d5670ff fix test_pick test time is too long (#16066)
add 5699939 [fix] Support nullop in `transpose` (#15865)
add 5def003 NumPy-compatible Mean, Std and Var (#16014)
add 1abf05b adding "total" (total time) to profiler aggregate stats
sorting criteria (#16055)
add a8ba6d9 fix flaky test (#16074)
add 692f3c4 Graph Partition API (#15886)
add 767e3f1 Add fluent methods mean, std, var for ndarray (#16077)
add f195098 fix some test files test time is too long (#16067)
add 4c72d27 new raise mode for nd.take and fix backward for wrap mode
(#15887)
add 5b301c6 Typedef cleanup (#15899)
add 9b9326f add KEY for Tao Lv (#16081)
add 65e37ca numpy multinomial op (#15878)
add 07b4470 MKL-DNN RNN checks NDArray version (#16071)
add c742ef1 add numpy operator remainder (#16080)
add 4333a7b Update readme and project.clj comment (#16084)
add 6122dfc Add Large tensor vector test cases (#15941)
add bc90e20 typo in docs (#16094)
add c6a92d9 remove 'foo' and other print msg from test (#16088)
add b7071c4 Enable tvm_op for ci (#15889)
add d0fa8c0 Test large vector mean operator and fix a few bugs (#16079)
add d60be31 Fix gradient tensor mutate in
`{adam/ftrl/rmprop/rmspropalex}_update`. (#15768)
add 7f57e8e [WIP] New Website: New Docs [1/3] (#15884)
new 5c5a619 Merge remote-tracking branch 'origin/master' into mkldnn-v1.0
The 1 revisions listed above as "new" are entirely new to this
repository and will be described in separate emails. The revisions
listed as "add" were already present in the repository and have only
been added to this reference.
Summary of changes:
3rdparty/mshadow/mshadow/cuda/tensor_gpu-inl.cuh | 26 +-
3rdparty/mshadow/mshadow/extension/slice.h |4 +-
3rdparty/mshadow/mshadow/tensor.h |4 +-
svn commit: r35639 - in /dev/incubator/mxnet/1.5.1.rc0: ./ apache-mxnet-src-1.5.1.rc0-incubating.tar.gz apache-mxnet-src-1.5.1.rc0-incubating.tar.gz.asc apache-mxnet-src-1.5.1.rc0-incubating.tar.gz.sh
Author: taolv Date: Fri Sep 6 13:55:31 2019 New Revision: 35639 Log: Add mxnet-1.5.1.rc0 Added: dev/incubator/mxnet/1.5.1.rc0/ dev/incubator/mxnet/1.5.1.rc0/apache-mxnet-src-1.5.1.rc0-incubating.tar.gz (with props) dev/incubator/mxnet/1.5.1.rc0/apache-mxnet-src-1.5.1.rc0-incubating.tar.gz.asc (with props) dev/incubator/mxnet/1.5.1.rc0/apache-mxnet-src-1.5.1.rc0-incubating.tar.gz.sha512 Added: dev/incubator/mxnet/1.5.1.rc0/apache-mxnet-src-1.5.1.rc0-incubating.tar.gz == Binary file - no diff available. Propchange: dev/incubator/mxnet/1.5.1.rc0/apache-mxnet-src-1.5.1.rc0-incubating.tar.gz -- svn:mime-type = application/x-gzip Added: dev/incubator/mxnet/1.5.1.rc0/apache-mxnet-src-1.5.1.rc0-incubating.tar.gz.asc == Binary file - no diff available. Propchange: dev/incubator/mxnet/1.5.1.rc0/apache-mxnet-src-1.5.1.rc0-incubating.tar.gz.asc -- svn:mime-type = application/pgp-signature Added: dev/incubator/mxnet/1.5.1.rc0/apache-mxnet-src-1.5.1.rc0-incubating.tar.gz.sha512 == --- dev/incubator/mxnet/1.5.1.rc0/apache-mxnet-src-1.5.1.rc0-incubating.tar.gz.sha512 (added) +++ dev/incubator/mxnet/1.5.1.rc0/apache-mxnet-src-1.5.1.rc0-incubating.tar.gz.sha512 Fri Sep 6 13:55:31 2019 @@ -0,0 +1 @@ +c0eab8e8728112d1c7b964934c9a665c4ca0336cf6e58f0c9c268ca0812173c78b0a94e6e173a23a06e0b53d15995f81b51b238eb167150455be22244b6c1a42 apache-mxnet-src-1.5.1.rc0-incubating.tar.gz
[incubator-mxnet-site] branch asf-site updated: Bump the publish timestamp.
This is an automated email from the ASF dual-hosted git repository. marcoabreu pushed a commit to branch asf-site in repository https://gitbox.apache.org/repos/asf/incubator-mxnet-site.git The following commit(s) were added to refs/heads/asf-site by this push: new d6ccb29 Bump the publish timestamp. d6ccb29 is described below commit d6ccb29a2dc770801d3e55c7386206cf760026ce Author: mxnet-ci AuthorDate: Fri Sep 6 13:33:14 2019 + Bump the publish timestamp. --- date.txt | 1 + 1 file changed, 1 insertion(+) diff --git a/date.txt b/date.txt new file mode 100644 index 000..e48d7bf --- /dev/null +++ b/date.txt @@ -0,0 +1 @@ +Fri Sep 6 13:33:14 UTC 2019
[GitHub] [incubator-mxnet] haojin2 commented on a change in pull request #15984: [DO NOT REVIEW] [DO NOT MERGE] General reduce compute for tvm ops and TVM version of sum
haojin2 commented on a change in pull request #15984: [DO NOT REVIEW] [DO NOT
MERGE] General reduce compute for tvm ops and TVM version of sum
URL: https://github.com/apache/incubator-mxnet/pull/15984#discussion_r321724004
##
File path: contrib/tvmop/basic/reduce.py
##
@@ -0,0 +1,96 @@
+# Licensed to the Apache Software Foundation (ASF) under one
+# or more contributor license agreements. See the NOTICE file
+# distributed with this work for additional information
+# regarding copyright ownership. The ASF licenses this file
+# to you under the Apache License, Version 2.0 (the
+# "License"); you may not use this file except in compliance
+# with the License. You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing,
+# software distributed under the License is distributed on an
+# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+# KIND, either express or implied. See the License for the
+# specific language governing permissions and limitations
+# under the License.
+
+# coding: utf-8
+import tvm
+from .. import defop, AllTypes, RealTypes
+from .. import assign_by_req, reduce_axes
+# AllTypes = ["float32", "float64", "float16", "uint8", "int8", "int32",
"int64"]
+# RealTypes = ["float32", "float64", "float16"]
+# AccTypes = {'float16': 'float32', 'float32': 'float64', 'float64': 'float64'}
+
+
+# def assign_by_req(a, req, otype):
+# b = tvm.placeholder(a.shape, name='assign_by_req_b', dtype=otype)
+# if (req == "kAddTo"):
+# c = tvm.compute(a.shape, lambda *idx: a[idx].astype(otype) + b[idx])
+# else:
+# c = tvm.compute(a.shape, lambda *idx: a[idx].astype(otype))
+# return b, c
+#
+# def reduce_axes(X, axes, reducer, atype=None):
+# def get_index(idx, ridx):
+# j = 0
+# k = 0
+# ret = []
+# for val in axes:
+# ret.append(idx[j] if val == 0 else ridx[k])
+# j += (val == 0)
+# k += (val != 0)
+# return tuple(ret)
+#
+# ishape = X.shape
+# odim = (len(ishape) + 1 - axes[0]) // 2
+# oshape = [tvm.var('odim.%d' % i, 'int32') for i in range(odim)]
+# if atype is None:
+# atype = X.dtype
+# ridx = [tvm.reduce_axis((0, ishape[i]), name='r%d' % i) for (i, val) in
enumerate(axes) if val == 1]
+# ret = tvm.compute(oshape, lambda *idx: reducer(X[get_index(idx,
ridx)].astype(atype), axis=ridx), name='ret')
+# return ret
+
+def compute_reduce(dtype, otype, reducer, initial, ndim, reduce1st, req):
Review comment:
Okay
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
With regards,
Apache Git Services
[GitHub] [incubator-mxnet] kshitij12345 commented on issue #15531: [MXNET-978] Higher Order Gradient Support `arctan`, `arctanh`, `radians`.
kshitij12345 commented on issue #15531: [MXNET-978] Higher Order Gradient Support `arctan`, `arctanh`, `radians`. URL: https://github.com/apache/incubator-mxnet/pull/15531#issuecomment-528842012 @apeforest @larroy Gentle ping. Could you please review it again? This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] [incubator-mxnet] QueensGambit closed issue #15640: Performance regression for MXNet 1.5.0
QueensGambit closed issue #15640: Performance regression for MXNet 1.5.0 URL: https://github.com/apache/incubator-mxnet/issues/15640 This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] [incubator-mxnet] QueensGambit commented on issue #15640: Performance regression for MXNet 1.5.0
QueensGambit commented on issue #15640: Performance regression for MXNet 1.5.0 URL: https://github.com/apache/incubator-mxnet/issues/15640#issuecomment-528804630 Great news! I'm happy to update to MXNet-1.6.0 or MXNet-1.6.0-Dev for the next _CrazyAra_ release. Thank you for awesome work. This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] [incubator-mxnet] ElaineBao commented on issue #15640: Performance regression for MXNet 1.5.0
ElaineBao commented on issue #15640: Performance regression for MXNet 1.5.0 URL: https://github.com/apache/incubator-mxnet/issues/15640#issuecomment-528768295 Hi, @QueensGambit, this issue has been fixed in mxnet 1.6. As version of 1.6 has not released yet, you can install nightly build version: `pip install mxnet-mkl==1.6.0b20190903` . I've tested above scripts on this version: ``` op: relu shape:all, total_count:38500, total_time:314.6143236053, avg_time:0.008171800613246741 shape:mb1ic128, total_count:5500, total_time:60.11501309865, avg_time:0.01093000238158 shape:mb1ic256ih8iw8, total_count:4400, total_time:32.9472733746, avg_time:0.0074880166750001045 shape:mb1ic128ih8iw8, total_count:2200, total_time:13.95386323001, avg_time:0.0063426651045454556 ... op: sigmoid total_count:5500, total_time:62.9402109975, avg_time:0.011443674725454682 shape:mb1ic256, total_count:5500, total_time:62.9402109975, avg_time:0.011443674725454682 ``` As you can see, relu with shape `mb1ic128` and sigmoid with shape `mb1ic256` have a faster performance than that in 1.5.0, and achieve the same result as 1.4.1. This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[incubator-mxnet-site] branch asf-site updated: Bump the publish timestamp.
This is an automated email from the ASF dual-hosted git repository. marcoabreu pushed a commit to branch asf-site in repository https://gitbox.apache.org/repos/asf/incubator-mxnet-site.git The following commit(s) were added to refs/heads/asf-site by this push: new 673b07a Bump the publish timestamp. 673b07a is described below commit 673b07ac8fb488f09430aaacd2c26ce0868701a4 Author: mxnet-ci AuthorDate: Fri Sep 6 07:40:22 2019 + Bump the publish timestamp. --- date.txt | 1 + 1 file changed, 1 insertion(+) diff --git a/date.txt b/date.txt new file mode 100644 index 000..1ee39b0 --- /dev/null +++ b/date.txt @@ -0,0 +1 @@ +Fri Sep 6 07:40:22 UTC 2019
[GitHub] [incubator-mxnet] mxnet-label-bot commented on issue #16110: ndarray treated uint8 as signed value
mxnet-label-bot commented on issue #16110: ndarray treated uint8 as signed value URL: https://github.com/apache/incubator-mxnet/issues/16110#issuecomment-528728517 Hey, this is the MXNet Label Bot. Thank you for submitting the issue! I will try and suggest some labels so that the appropriate MXNet community members can help resolve it. Here are my recommended label(s): Bug This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] [incubator-mxnet] dwSun opened a new issue #16110: ndarray treated uint8 as signed value
dwSun opened a new issue #16110: ndarray treated uint8 as signed value
URL: https://github.com/apache/incubator-mxnet/issues/16110
## Description
ndarray treated uint8 as signed value, cause nd.mean and nd.sum return
confused values.
## Environment info (Required)
```
--Python Info--
Version : 3.7.4+
Compiler : GCC 9.2.1 20190827
Build: ('default', 'Sep 4 2019 08:03:05')
Arch : ('64bit', 'ELF')
Pip Info---
Version : 19.2.3
Directory: /home/david/.local/lib/python3.7/site-packages/pip
--MXNet Info---
Version : 1.5.0
Directory: /home/david/.local/lib/python3.7/site-packages/mxnet
Commit Hash : 75a9e187d00a8b7ebc71412a02ed0e3ae489d91f
Library :
['/home/david/.local/lib/python3.7/site-packages/mxnet/libmxnet.so']
Build features:
✖ CUDA
✖ CUDNN
✖ NCCL
✖ CUDA_RTC
✖ TENSORRT
✔ CPU_SSE
✔ CPU_SSE2
✔ CPU_SSE3
✔ CPU_SSE4_1
✔ CPU_SSE4_2
✖ CPU_SSE4A
✔ CPU_AVX
✖ CPU_AVX2
✖ OPENMP
✖ SSE
✔ F16C
✖ JEMALLOC
✖ BLAS_OPEN
✖ BLAS_ATLAS
✖ BLAS_MKL
✖ BLAS_APPLE
✔ LAPACK
✔ MKLDNN
✔ OPENCV
✖ CAFFE
✖ PROFILER
✔ DIST_KVSTORE
✖ CXX14
✖ INT64_TENSOR_SIZE
✔ SIGNAL_HANDLER
✖ DEBUG
--System Info--
Platform : Linux-5.2.0-2-amd64-x86_64-with-debian-bullseye-sid
system : Linux
node : Zarus
release : 5.2.0-2-amd64
version : #1 SMP Debian 5.2.9-2 (2019-08-21)
--Hardware Info--
machine : x86_64
processor:
Architecture:x86_64
CPU op-mode(s): 32-bit, 64-bit
Byte Order: Little Endian
Address sizes: 39 bits physical, 48 bits virtual
CPU(s): 4
On-line CPU(s) list: 0-3
Thread(s) per core: 2
Core(s) per socket: 2
Socket(s): 1
NUMA node(s):1
Vendor ID: GenuineIntel
CPU family: 6
Model: 78
Model name: Intel(R) Core(TM) i7-6500U CPU @ 2.50GHz
Stepping:3
CPU MHz: 2700.094
CPU max MHz: 3100.
CPU min MHz: 400.
BogoMIPS:5184.00
Virtualization: VT-x
L1d cache: 64 KiB
L1i cache: 64 KiB
L2 cache:512 KiB
L3 cache:4 MiB
NUMA node0 CPU(s): 0-3
Vulnerability L1tf: Mitigation; PTE Inversion; VMX conditional
cache flushes, SMT vulnerable
Vulnerability Mds: Mitigation; Clear CPU buffers; SMT vulnerab
le
Vulnerability Meltdown: Mitigation; PTI
Vulnerability Spec store bypass: Mitigation; Speculative Store Bypass disabl
ed via prctl and seccomp
Vulnerability Spectre v1:Mitigation; usercopy/swapgs barriers and __
user pointer sanitization
Vulnerability Spectre v2:Mitigation; Full generic retpoline, IBPB co
nditional, IBRS_FW, STIBP conditional, RSB
filling
Flags: fpu vme de pse tsc msr pae mce cx8 apic sep
mtrr pge mca cmov pat pse36 clflush dts ac
pi mmx fxsr sse sse2 ss ht tm pbe syscall n
x pdpe1gb rdtscp lm constant_tsc art arch_p
erfmon pebs bts rep_good nopl xtopology non
stop_tsc cpuid aperfmperf tsc_known_freq pn
i pclmulqdq dtes64 monitor ds_cpl vmx est t
m2 ssse3 sdbg fma cx16 xtpr pdcm pcid sse4_
1 sse4_2 x2apic movbe popcnt tsc_deadline_t
imer aes xsave avx f16c rdrand lahf_lm abm
3dnowprefetch cpuid_fault epb invpcid_singl
e pti ssbd ibrs ibpb stibp tpr_shadow vnmi
flexpriority ept vpid ept_ad fsgsbase tsc_a
djust bmi1 avx2 smep bmi2 erms invpcid mpx
rdseed adx smap clflushopt intel_pt xsaveop
t xsavec xgetbv1 xsaves dtherm ida arat pln
pts hwp hwp_notify hwp_act_window hwp_epp
md_clear flush_l1d
[GitHub] [incubator-mxnet] xidulu opened a new pull request #16109: [Numpy] Numpy behavior normal distribution
xidulu opened a new pull request #16109: [Numpy] Numpy behavior normal distribution URL: https://github.com/apache/incubator-mxnet/pull/16109 ## Description ## As title ## Checklist ## ### Essentials ### Please feel free to remove inapplicable items for your PR. - [x] The PR title starts with [MXNET-$JIRA_ID], where $JIRA_ID refers to the relevant [JIRA issue](https://issues.apache.org/jira/projects/MXNET/issues) created (except PRs with tiny changes) - [x] Changes are complete (i.e. I finished coding on this PR) - [x] All changes have test coverage: - Unit tests are added for small changes to verify correctness (e.g. adding a new operator) - Nightly tests are added for complicated/long-running ones (e.g. changing distributed kvstore) - Build tests will be added for build configuration changes (e.g. adding a new build option with NCCL) - [x] Code is well-documented: - For user-facing API changes, API doc string has been updated. - For new C++ functions in header files, their functionalities and arguments are documented. - For new examples, README.md is added to explain the what the example does, the source of the dataset, expected performance on test set and reference to the original paper if applicable - Check the API doc at http://mxnet-ci-doc.s3-accelerate.dualstack.amazonaws.com/PR-$PR_ID/$BUILD_ID/index.html - [x] To the my best knowledge, examples are either not affected by this change, or have been fixed to be compatible with this change ## Comments ## - If this change is a backward incompatible change, why must this change be made. - Interesting edge cases to note here This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
