[incubator-mxnet-site] branch asf-site updated: Bump the publish timestamp.
This is an automated email from the ASF dual-hosted git repository. zhasheng 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 1c2245b Bump the publish timestamp. 1c2245b is described below commit 1c2245b10b2a324d57c5399290888921d46aa851 Author: mxnet-ci AuthorDate: Sun Feb 3 07:08:57 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..4f4d489 --- /dev/null +++ b/date.txt @@ -0,0 +1 @@ +Sun Feb 3 07:08:57 UTC 2019
[GitHub] zhreshold commented on a change in pull request #14031: Fix transposed convolution in CPU w/o MKLDNN.
zhreshold commented on a change in pull request #14031: Fix transposed convolution in CPU w/o MKLDNN. URL: https://github.com/apache/incubator-mxnet/pull/14031#discussion_r253289470 ## File path: tests/python/unittest/test_gluon.py ## @@ -503,6 +503,40 @@ def test_deconv(): # layer = nn.Conv3DTranspose(16, (3, 3, 3), layout='NDHWC', in_channels=4) # # check_layer_forward(layer, (1, 10, 10, 10, 4)) +@with_seed() +def test_deconv_dilation(): Review comment: Since deconv is a really important OP, I suggest to visit the original deconv test cases and add dilation > 1 cases alongside the old tests. This ensures better coverage than this single test case. Feel free to keep this unittest which LGTM as well. This is an automated message from the Apache Git Service. To respond to the message, please log on 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] KellenSunderland commented on issue #14040: Reformat of TensorRT to use subgraph API
KellenSunderland commented on issue #14040: Reformat of TensorRT to use subgraph API URL: https://github.com/apache/incubator-mxnet/pull/14040#issuecomment-460027375 ONNX-Tensorrt might fail to build as it adds a new default target that requires a header not in a standard search path. There's a few ways to work around the problem, we could just build the library and not that tool in CI, or we could include the header folder location in the search path like so: https://github.com/apache/incubator-mxnet/pull/13906/files#diff-56133c25b5a238b76f54c0928f05a8e6 It should allow the TensorRT build to pass CI if we make that change in this PR. This is an automated message from the Apache Git Service. To respond to the message, please log on 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] KellenSunderland removed a comment on issue #13906: [MXNET-703] Update onnx-tensorrt for int8/fp16 support
KellenSunderland removed a comment on issue #13906: [MXNET-703] Update onnx-tensorrt for int8/fp16 support URL: https://github.com/apache/incubator-mxnet/pull/13906#issuecomment-460020295 18 days, no review, closing. This is an automated message from the Apache Git Service. To respond to the message, please log on 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] KellenSunderland opened a new pull request #13906: [MXNET-703] Update onnx-tensorrt for int8/fp16 support
KellenSunderland opened a new pull request #13906: [MXNET-703] Update onnx-tensorrt for int8/fp16 support URL: https://github.com/apache/incubator-mxnet/pull/13906 ## Description ## Update onnx-tensorrt with new support for different data types, including (importantly for v100) fp16. This will not fully enable int8/fp16 inference yet, but lays the groundwork for further updates. ## 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 This is an automated message from the Apache Git Service. To respond to the message, please log on 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] KellenSunderland commented on a change in pull request #14040: Reformat of TensorRT to use subgraph API
KellenSunderland commented on a change in pull request #14040: Reformat of
TensorRT to use subgraph API
URL: https://github.com/apache/incubator-mxnet/pull/14040#discussion_r253288751
##
File path: src/operator/subgraph/tensorrt/tensorrt-inl.h
##
@@ -0,0 +1,217 @@
+#ifndef MXNET_OPERATOR_SUBGRAPH_TENSORRT_TENSORRT_INL_H_
+#define MXNET_OPERATOR_SUBGRAPH_TENSORRT_TENSORRT_INL_H_
+/*
+ * 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) 2018 by Contributors
+ * \file tensorrt-inl.h
+ * \brief TensorRT operation registration
+ * \author Marek Kolodziej, Clement Fuji Tsang
+*/
+
+#if MXNET_USE_TENSORRT
+
+
+#include "../common.h"
+#include "../subgraph_property.h"
+#include "nnvm_to_onnx-inl.h"
+#include "./onnx_to_tensorrt.h"
+
+namespace mxnet {
+namespace op {
+
+using int64 = ::google::protobuf::int64;
+
+struct TRTParam {
+ std::unordered_map inputs_to_idx;
+ std::unordered_map outputs_to_idx;
+ std::unordered_map params_map;
+};
+
+struct TRTEngineParam {
+ nvinfer1::IExecutionContext* trt_executor = nullptr;
+ std::vector > binding_vec;
+};
+
+
+class TensorrtSelector : public SubgraphSelector {
+ public:
+ const std::unordered_set unconditionalTRTops = {
+"Convolution",
+"BatchNorm",
+"elemwise_add",
+"elemwise_sub",
+"elemwise_mul",
+"rsqrt",
+"pad",
+"Pad",
+"mean",
+"FullyConnected",
+"Flatten",
+"SoftmaxOutput",
+ };
+
+ const std::unordered_set withWeightsOps = {
+"Convolution",
+"BatchNorm",
+"FullyConnected"
+ };
+
+ bool isTRTCompatible(const nnvm::Node &n) {
+const std::string op_name = n.op()->name;
+if (op_name == "Pooling") {
+ return (n.attrs.dict.at("pool_type") == "avg" ||
+ n.attrs.dict.at("pool_type") == "max");
+}
+
+if (unconditionalTRTops.count(op_name)) {
+ return true;
+}
+
+if (op_name == "Activation") {
+ return n.attrs.dict.at("act_type") == "relu" ||
+n.attrs.dict.at("act_type") == "tanh" ||
+n.attrs.dict.at("act_type") == "sigmoid";
+}
+
+return false;
+ }
+
+ bool Select(const nnvm::Node &n) override {
+return !n.is_variable() && isTRTCompatible(n);
+ }
+
+ bool SelectInput(const nnvm::Node &n, const nnvm::Node &new_node) override {
+if (new_node.is_variable()) {
+ if (withWeightsOps.count(n.op()->name)) {
+return n.inputs[0].node->attrs.name != new_node.attrs.name;
+ } else {
+return false;
+ }
+}
+if (isTRTCompatible(new_node))
+ return true;
+return false;
+ }
+
+ bool SelectOutput(const nnvm::Node &n, const nnvm::Node &new_node) override {
+ return isTRTCompatible(new_node);
+ }
+
+ std::vector Filter(const std::vector& candidates)
override {
+bool found_one = false;
+// TensorRT is interesting with at least 2 operations
+for (auto& n : candidates) {
+ if (!n->is_variable()) {
+if (found_one) {
+ return candidates;
+} else {
+ found_one = true;
+}
+ }
+}
+return std::vector();
+ }
+};
+
+class TensorrtProperty : public SubgraphProperty {
+ public:
+ static SubgraphPropertyPtr Create() {
+return std::make_shared();
+ }
+
+ nnvm::NodePtr CreateSubgraphNode(const nnvm::Symbol &sym,
+ const int subgraph_id = 0) const override {
Review comment:
If possible can we avoid default values in overriding functions.
This is an automated message from the Apache Git Service.
To respond to the message, please log on 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] KellenSunderland commented on a change in pull request #14040: Reformat of TensorRT to use subgraph API
KellenSunderland commented on a change in pull request #14040: Reformat of
TensorRT to use subgraph API
URL: https://github.com/apache/incubator-mxnet/pull/14040#discussion_r253288427
##
File path: src/operator/subgraph/tensorrt/tensorrt-inl.h
##
@@ -0,0 +1,217 @@
+#ifndef MXNET_OPERATOR_SUBGRAPH_TENSORRT_TENSORRT_INL_H_
+#define MXNET_OPERATOR_SUBGRAPH_TENSORRT_TENSORRT_INL_H_
+/*
+ * 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) 2018 by Contributors
+ * \file tensorrt-inl.h
+ * \brief TensorRT operation registration
+ * \author Marek Kolodziej, Clement Fuji Tsang
+*/
+
+#if MXNET_USE_TENSORRT
+
+
+#include "../common.h"
+#include "../subgraph_property.h"
+#include "nnvm_to_onnx-inl.h"
+#include "./onnx_to_tensorrt.h"
+
+namespace mxnet {
+namespace op {
+
+using int64 = ::google::protobuf::int64;
+
+struct TRTParam {
+ std::unordered_map inputs_to_idx;
+ std::unordered_map outputs_to_idx;
+ std::unordered_map params_map;
+};
+
+struct TRTEngineParam {
+ nvinfer1::IExecutionContext* trt_executor = nullptr;
+ std::vector > binding_vec;
+};
+
+
+class TensorrtSelector : public SubgraphSelector {
+ public:
+ const std::unordered_set unconditionalTRTops = {
+"Convolution",
+"BatchNorm",
+"elemwise_add",
+"elemwise_sub",
+"elemwise_mul",
+"rsqrt",
+"pad",
+"Pad",
+"mean",
+"FullyConnected",
+"Flatten",
+"SoftmaxOutput",
+ };
+
+ const std::unordered_set withWeightsOps = {
+"Convolution",
+"BatchNorm",
+"FullyConnected"
+ };
+
+ bool isTRTCompatible(const nnvm::Node &n) {
+const std::string op_name = n.op()->name;
+if (op_name == "Pooling") {
+ return (n.attrs.dict.at("pool_type") == "avg" ||
+ n.attrs.dict.at("pool_type") == "max");
+}
+
+if (unconditionalTRTops.count(op_name)) {
+ return true;
+}
+
+if (op_name == "Activation") {
+ return n.attrs.dict.at("act_type") == "relu" ||
+n.attrs.dict.at("act_type") == "tanh" ||
+n.attrs.dict.at("act_type") == "sigmoid";
+}
+
+return false;
+ }
+
+ bool Select(const nnvm::Node &n) override {
+return !n.is_variable() && isTRTCompatible(n);
+ }
+
+ bool SelectInput(const nnvm::Node &n, const nnvm::Node &new_node) override {
+if (new_node.is_variable()) {
+ if (withWeightsOps.count(n.op()->name)) {
+return n.inputs[0].node->attrs.name != new_node.attrs.name;
+ } else {
+return false;
+ }
+}
+if (isTRTCompatible(new_node))
+ return true;
+return false;
+ }
+
+ bool SelectOutput(const nnvm::Node &n, const nnvm::Node &new_node) override {
+ return isTRTCompatible(new_node);
+ }
+
+ std::vector Filter(const std::vector& candidates)
override {
+bool found_one = false;
+// TensorRT is interesting with at least 2 operations
+for (auto& n : candidates) {
+ if (!n->is_variable()) {
+if (found_one) {
+ return candidates;
+} else {
+ found_one = true;
+}
+ }
+}
+return std::vector();
+ }
+};
+
+class TensorrtProperty : public SubgraphProperty {
+ public:
+ static SubgraphPropertyPtr Create() {
+return std::make_shared();
+ }
+
+ nnvm::NodePtr CreateSubgraphNode(const nnvm::Symbol &sym,
+ const int subgraph_id = 0) const override {
+nnvm::NodePtr n = nnvm::Node::Create();
+nnvm::Symbol new_sym;
+std::unique_copy(sym.outputs.begin(), sym.outputs.end(),
+std::back_inserter(new_sym.outputs), [](
+nnvm::NodeEntry lhs, nnvm::NodeEntry rhs) {
+ return lhs.index == rhs.index && lhs.node.get() == rhs.node.get();
+});
+n->attrs.name = "TensorRT" + std::to_string(subgraph_id);
+n->attrs.op = Op::Get("_TensorRT");
+CHECK(n->attrs.op);
+n->attrs.subgraphs.emplace_back(std::make_shared(new_sym));
+std::ostringstream params_oss;
+for (auto &e : new_sym.ListInputNames(nnvm::Symbol::kAll)) {
+ params_oss << e << ";";
+}
+auto tensorrt_params_names = params_oss.str();
+tensorrt_params_names.pop_back();
+n->attrs.dict["subgraph_params_names"] = tensorrt_params_names;
+TRTParam param;
[GitHub] KellenSunderland commented on a change in pull request #14040: Reformat of TensorRT to use subgraph API
KellenSunderland commented on a change in pull request #14040: Reformat of
TensorRT to use subgraph API
URL: https://github.com/apache/incubator-mxnet/pull/14040#discussion_r253288413
##
File path: src/operator/subgraph/tensorrt/tensorrt-inl.h
##
@@ -0,0 +1,217 @@
+#ifndef MXNET_OPERATOR_SUBGRAPH_TENSORRT_TENSORRT_INL_H_
+#define MXNET_OPERATOR_SUBGRAPH_TENSORRT_TENSORRT_INL_H_
+/*
+ * 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) 2018 by Contributors
+ * \file tensorrt-inl.h
+ * \brief TensorRT operation registration
+ * \author Marek Kolodziej, Clement Fuji Tsang
+*/
+
+#if MXNET_USE_TENSORRT
+
+
+#include "../common.h"
+#include "../subgraph_property.h"
+#include "nnvm_to_onnx-inl.h"
+#include "./onnx_to_tensorrt.h"
+
+namespace mxnet {
+namespace op {
+
+using int64 = ::google::protobuf::int64;
+
+struct TRTParam {
+ std::unordered_map inputs_to_idx;
+ std::unordered_map outputs_to_idx;
+ std::unordered_map params_map;
+};
+
+struct TRTEngineParam {
+ nvinfer1::IExecutionContext* trt_executor = nullptr;
+ std::vector > binding_vec;
+};
+
+
+class TensorrtSelector : public SubgraphSelector {
+ public:
+ const std::unordered_set unconditionalTRTops = {
+"Convolution",
+"BatchNorm",
+"elemwise_add",
+"elemwise_sub",
+"elemwise_mul",
+"rsqrt",
+"pad",
+"Pad",
+"mean",
+"FullyConnected",
+"Flatten",
+"SoftmaxOutput",
+ };
+
+ const std::unordered_set withWeightsOps = {
+"Convolution",
+"BatchNorm",
+"FullyConnected"
+ };
+
+ bool isTRTCompatible(const nnvm::Node &n) {
+const std::string op_name = n.op()->name;
+if (op_name == "Pooling") {
+ return (n.attrs.dict.at("pool_type") == "avg" ||
+ n.attrs.dict.at("pool_type") == "max");
+}
+
+if (unconditionalTRTops.count(op_name)) {
+ return true;
+}
+
+if (op_name == "Activation") {
+ return n.attrs.dict.at("act_type") == "relu" ||
+n.attrs.dict.at("act_type") == "tanh" ||
+n.attrs.dict.at("act_type") == "sigmoid";
+}
+
+return false;
+ }
+
+ bool Select(const nnvm::Node &n) override {
+return !n.is_variable() && isTRTCompatible(n);
+ }
+
+ bool SelectInput(const nnvm::Node &n, const nnvm::Node &new_node) override {
+if (new_node.is_variable()) {
+ if (withWeightsOps.count(n.op()->name)) {
+return n.inputs[0].node->attrs.name != new_node.attrs.name;
+ } else {
+return false;
+ }
+}
+if (isTRTCompatible(new_node))
Review comment:
Can be simplified to
```C++
return isTRTCompatible(new_node);
```
This is an automated message from the Apache Git Service.
To respond to the message, please log on 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] KellenSunderland commented on a change in pull request #14040: Reformat of TensorRT to use subgraph API
KellenSunderland commented on a change in pull request #14040: Reformat of
TensorRT to use subgraph API
URL: https://github.com/apache/incubator-mxnet/pull/14040#discussion_r253288287
##
File path: src/operator/subgraph/tensorrt/onnx_to_tensorrt.cc
##
@@ -127,16 +128,15 @@ nvinfer1::ICudaEngine* onnxToTrtCtx(
}
throw dmlc::Error("Cannot parse ONNX into TensorRT Engine");
}
-
- bool fp16 = trt_builder->platformHasFastFp16();
-
+ if (dmlc::GetEnv("MXNET_TENSORRT_USE_FP16", true)) {
+if (trt_builder->platformHasFastFp16()) {
+ trt_builder->setFp16Mode(true);
+} else {
+ LOG(INFO) << "WARNING: TensorRT can't use fp16 on this plateform";
Review comment:
Also we're logging INFO level logs but have WARNING in the message. I'd
remove the warning from the message and set log level to warning. (this is a
common issue in our codebase).
This is an automated message from the Apache Git Service.
To respond to the message, please log on 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] KellenSunderland commented on a change in pull request #14040: Reformat of TensorRT to use subgraph API
KellenSunderland commented on a change in pull request #14040: Reformat of
TensorRT to use subgraph API
URL: https://github.com/apache/incubator-mxnet/pull/14040#discussion_r253288274
##
File path: src/operator/subgraph/tensorrt/onnx_to_tensorrt.cc
##
@@ -127,16 +128,15 @@ nvinfer1::ICudaEngine* onnxToTrtCtx(
}
throw dmlc::Error("Cannot parse ONNX into TensorRT Engine");
}
-
- bool fp16 = trt_builder->platformHasFastFp16();
-
+ if (dmlc::GetEnv("MXNET_TENSORRT_USE_FP16", true)) {
+if (trt_builder->platformHasFastFp16()) {
+ trt_builder->setFp16Mode(true);
+} else {
+ LOG(INFO) << "WARNING: TensorRT can't use fp16 on this plateform";
Review comment:
plateform -> platform
This is an automated message from the Apache Git Service.
To respond to the message, please log on 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] KellenSunderland commented on a change in pull request #14040: Reformat of TensorRT to use subgraph API
KellenSunderland commented on a change in pull request #14040: Reformat of
TensorRT to use subgraph API
URL: https://github.com/apache/incubator-mxnet/pull/14040#discussion_r253288158
##
File path: src/operator/subgraph/tensorrt/nnvm_to_onnx.cc
##
@@ -411,6 +383,49 @@ void ConvertElementwiseAdd(NodeProto* node_proto, const
NodeAttrs& /*attrs*/,
node_proto->set_op_type("Add");
}
+inline TensorProto_DataType ConvertDType(int dtype) {
+ switch (dtype) {
+case mshadow::kFloat64:
+ return TensorProto_DataType_DOUBLE;
+case mshadow::kFloat32:
+ return TensorProto_DataType_FLOAT;
+case mshadow::kFloat16:
+ return TensorProto_DataType_FLOAT16;
+case mshadow::kUint8:
+ return TensorProto_DataType_UINT8;
+case mshadow::kInt32:
+ return TensorProto_DataType_INT32;
+case mshadow::kInt8:
+ return TensorProto_DataType_INT8;
+case mshadow::kInt64:
+ return TensorProto_DataType_INT64;
+default:
+ return TensorProto_DataType_UNDEFINED;
+ }
+}
+
+inline std::string StringDType(int dtype) {
Review comment:
Is this still needed, don't see any references to it.
This is an automated message from the Apache Git Service.
To respond to the message, please log on 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] KellenSunderland commented on a change in pull request #14040: Reformat of TensorRT to use subgraph API
KellenSunderland commented on a change in pull request #14040: Reformat of TensorRT to use subgraph API URL: https://github.com/apache/incubator-mxnet/pull/14040#discussion_r253288021 ## File path: include/mxnet/c_api.h ## @@ -1835,12 +1835,6 @@ MXNET_DLL int MXExecutorReshape(int partial_shaping, ExecutorHandle shared_exec, ExecutorHandle *out); -/*! - * \brief get optimized graph from graph executor - */ -MXNET_DLL int MXExecutorGetOptimizedSymbol(ExecutorHandle handle, - SymbolHandle *out); - Review comment: It affect semantic versioning if we remove it (it can break compilation on downstream projects) so if there's not a strong reason to remove it we should leave it in for that reason. Can it still preform a useful function (for example showing the graph after optimization)? This is an automated message from the Apache Git Service. To respond to the message, please log on 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] KellenSunderland commented on a change in pull request #14040: Reformat of TensorRT to use subgraph API
KellenSunderland commented on a change in pull request #14040: Reformat of TensorRT to use subgraph API URL: https://github.com/apache/incubator-mxnet/pull/14040#discussion_r253288021 ## File path: include/mxnet/c_api.h ## @@ -1835,12 +1835,6 @@ MXNET_DLL int MXExecutorReshape(int partial_shaping, ExecutorHandle shared_exec, ExecutorHandle *out); -/*! - * \brief get optimized graph from graph executor - */ -MXNET_DLL int MXExecutorGetOptimizedSymbol(ExecutorHandle handle, - SymbolHandle *out); - Review comment: It affects semantic versioning if we remove it (it can break compilation on downstream projects) so if there's not a strong reason to remove it we should leave it in for that reason. Can it still preform a useful function (for example showing the graph after optimization)? This is an automated message from the Apache Git Service. To respond to the message, please log on 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] Caenorst commented on a change in pull request #14040: Reformat of TensorRT to use subgraph API
Caenorst commented on a change in pull request #14040: Reformat of TensorRT to use subgraph API URL: https://github.com/apache/incubator-mxnet/pull/14040#discussion_r253287088 ## File path: include/mxnet/c_api.h ## @@ -1835,12 +1835,6 @@ MXNET_DLL int MXExecutorReshape(int partial_shaping, ExecutorHandle shared_exec, ExecutorHandle *out); -/*! - * \brief get optimized graph from graph executor - */ -MXNET_DLL int MXExecutorGetOptimizedSymbol(ExecutorHandle handle, - SymbolHandle *out); - Review comment: I actually created this API just for the previous TensorRT implementation, I'm not sure it is used anywhere else. It could actually be used in case you are calling a subgraph backend with variable environment. Do you want to keep it ? This is an automated message from the Apache Git Service. To respond to the message, please log on 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] KellenSunderland closed pull request #13906: [MXNET-703] Update onnx-tensorrt for int8/fp16 support
KellenSunderland closed pull request #13906: [MXNET-703] Update onnx-tensorrt for int8/fp16 support URL: https://github.com/apache/incubator-mxnet/pull/13906 This is an automated message from the Apache Git Service. To respond to the message, please log on 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] KellenSunderland commented on issue #13906: [MXNET-703] Update onnx-tensorrt for int8/fp16 support
KellenSunderland commented on issue #13906: [MXNET-703] Update onnx-tensorrt for int8/fp16 support URL: https://github.com/apache/incubator-mxnet/pull/13906#issuecomment-460020295 18 days, no review, closing. This is an automated message from the Apache Git Service. To respond to the message, please log on 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] rajeshii opened a new pull request #14060: Exclude concat layer for gpu quantization
rajeshii opened a new pull request #14060: Exclude concat layer for gpu quantization URL: https://github.com/apache/incubator-mxnet/pull/14060 ## Description ## Exclude concat layer for gpu quantization since #13297 enable quantizaed_concat op for cpu. below is the error log before this fix: ``` python imagenet_inference.py --symbol-file=./model/imagenet1k-inception-bn-quantized-5batches-naive-symbol.json --param-file=./model/imagenet1k-inception-bn-quantized-.params --rgb-mean=123.68,116.779,103.939 --num-skipped-batches=50 --num-inference-batches=500 --dataset=./data/val_256_q90.rec INFO:logger:batch size = 32 for inference INFO:logger:rgb_mean = 123.68,116.779,103.939 INFO:logger:rgb_std = 1,1,1 INFO:logger:label_name = softmax_label INFO:logger:Input data shape = (3, 224, 224) INFO:logger:Dataset for inference: ./data/val_256_q90.rec [10:15:59] src/io/iter_image_recordio_2.cc:172: ImageRecordIOParser2: ./data/val_256_q90.rec, use 39 threads for decoding.. INFO:logger:Loading symbol from file /home/chenxiny/s8_conv/example/quantization/./model/imagenet1k-inception-bn-quantized-5batches-naive-symbol.json INFO:logger:Loading params from file /home/chenxiny/s8_conv/example/quantization/./model/imagenet1k-inception-bn-quantized-.params INFO:logger:Skipping the first 50 batches INFO:logger:Running model ./model/imagenet1k-inception-bn-quantized-5batches-naive-symbol.json for inference [10:16:04] src/executor/attach_op_execs_pass.cc:351: Neither FCompute nor FComputeEx registered _contrib_quantized_concat ``` ## Checklist ## ### Essentials ### Please feel free to remove inapplicable items for your PR. - [ ] The PR title starts with [MXNET-$JIRA_ID], where $JIRA_ID refers to the relevant [JIRA issue](https://issues.apache.org/jira/projects/MXNET/issues) created (except PRs with tiny changes) - [x] Changes are complete (i.e. I finished coding on this PR) - [x] All changes have test coverage: - Unit tests are added for small changes to verify correctness (e.g. adding a new operator) - Nightly tests are added for complicated/long-running ones (e.g. changing distributed kvstore) - Build tests will be added for build configuration changes (e.g. adding a new build option with NCCL) - [x] Code is well-documented: - For user-facing API changes, API doc string has been updated. - For new C++ functions in header files, their functionalities and arguments are documented. - For new examples, README.md is added to explain the what the example does, the source of the dataset, expected performance on test set and reference to the original paper if applicable - Check the API doc at http://mxnet-ci-doc.s3-accelerate.dualstack.amazonaws.com/PR-$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 ### 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 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] pengzhao-intel commented on issue #13668: Update MKL-DNN to fix the Dense layer issue
pengzhao-intel commented on issue #13668: Update MKL-DNN to fix the Dense layer issue URL: https://github.com/apache/incubator-mxnet/pull/13668#issuecomment-460016722 @zheng-da please help verify the fix. This is an automated message from the Apache Git Service. To respond to the message, please log on 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] pengzhao-intel commented on issue #13668: Update MKL-DNN to fix the Dense layer issue
pengzhao-intel commented on issue #13668: Update MKL-DNN to fix the Dense layer issue URL: https://github.com/apache/incubator-mxnet/pull/13668#issuecomment-460016667 We will update the latest MKLDNN CI before the merge. This is an automated message from the Apache Git Service. To respond to the message, please log on 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] ZhennanQin commented on issue #14052: flaky test: test_operator.test_depthwise_convolution
ZhennanQin commented on issue #14052: flaky test: test_operator.test_depthwise_convolution URL: https://github.com/apache/incubator-mxnet/issues/14052#issuecomment-460016527 @mseth10, Thanks for reporting this. Confirmed it's caused by mkldnn bug. The fix is already in mkldnn master, but not available in its latest release. We will upgrade mkldnn when its next release comes out. This is an automated message from the Apache Git Service. To respond to the message, please log on 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] zheng-da commented on a change in pull request #14040: Reformat of TensorRT to use subgraph API
zheng-da commented on a change in pull request #14040: Reformat of TensorRT to use subgraph API URL: https://github.com/apache/incubator-mxnet/pull/14040#discussion_r253284118 ## File path: include/mxnet/c_api.h ## @@ -1835,12 +1835,6 @@ MXNET_DLL int MXExecutorReshape(int partial_shaping, ExecutorHandle shared_exec, ExecutorHandle *out); -/*! - * \brief get optimized graph from graph executor - */ -MXNET_DLL int MXExecutorGetOptimizedSymbol(ExecutorHandle handle, - SymbolHandle *out); - Review comment: why do you need to remove an API? This is an automated message from the Apache Git Service. To respond to the message, please log on 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] huangzhiyuan commented on issue #14056: A better split-2D(SliceChannel) op forward kernel for CPU
huangzhiyuan commented on issue #14056: A better split-2D(SliceChannel) op forward kernel for CPU URL: https://github.com/apache/incubator-mxnet/pull/14056#issuecomment-460014063 Seems the PR submitted recently block in the same test case, I will rebase my code after CNY, ths! This is an automated message from the Apache Git Service. To respond to the message, please log on 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. zhasheng 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 c4de8b2 Bump the publish timestamp. c4de8b2 is described below commit c4de8b225f331ead9fc92f13b854eaa74f8caebc Author: mxnet-ci AuthorDate: Sun Feb 3 01:08:22 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..ab916b7 --- /dev/null +++ b/date.txt @@ -0,0 +1 @@ +Sun Feb 3 01:08:22 UTC 2019
[GitHub] pengzhao-intel commented on issue #13697: [MKLDNN] Enable signed int8 support for convolution.
pengzhao-intel commented on issue #13697: [MKLDNN] Enable signed int8 support for convolution. URL: https://github.com/apache/incubator-mxnet/pull/13697#issuecomment-460013352 @xinyu-intel could you help verify the GPU accuracy with this PR? This is an automated message from the Apache Git Service. To respond to the message, please log on 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] stephenrawls commented on issue #14053: in-place reshape ops
stephenrawls commented on issue #14053: in-place reshape ops URL: https://github.com/apache/incubator-mxnet/pull/14053#issuecomment-460005020 @szha -- Thanks for putting in this patch! I have a couple questions about the operator `expand_dims`, which I understand you are actually changing the python code to *not* use that operator, but I think still has relevance. (1) I originally discovered this because I was using the C API to call MXImperativeInvoke() using the expand_dims operator, and I noticed it was causing a copy. This fix only effects the Python version when operating on ndarrays, not users of the expand_dims operator. I see that there is a path in the operator that checks if it is an in-place operation. Presumably it uses that path if I pass an output array that is the same NDArrayHandle as the input array? But what if I still need the original input array handle, and I want to create a new output array handle with the expanded dim but still not make a copy? (2) In the issue I created you commented that: "For symbol (and thus the hybridized version), since in-place identity is possible it should not matter". Can you talk a little more about that? I assume you mean that in this case: ``` x_expanded = x.expand_dims(1) y = x_expanded + foo ``` The engine can figure out that x is not needed again, and can thus turn the expand_dims(1) into an in-place operation that doesn't make a copy? I'm not very familiar with how this part of the code works, so what happens if you had code that looked like this? ``` x_expanded = x.expand_dims(1) y = x_expanded + foo z = 2 * x ``` i.e. the code still makes a reference to the original x, and thus presumably the engine can't decide to use the in-place version of expand_dims in that case, right? So I guess my question is -- Does the ability for the Syblolic / hybridized engine to elide the copy depend on the code not referencing the un-expanded version of the array after calling expand_dims()? If so, it seems like there will still be some use cases where an unexpected copy is happening. This is an automated message from the Apache Git Service. To respond to the message, please log on 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] stephenrawls commented on issue #13680: [MXNET-1121] Example to demonstrate the inference workflow using RNN
stephenrawls commented on issue #13680: [MXNET-1121] Example to demonstrate the inference workflow using RNN URL: https://github.com/apache/incubator-mxnet/pull/13680#issuecomment-46730 Looks good to me, thanks for the example. Made a few small comments in the code about sharing executor memory, and extracting model output. It would probably be nice to showcase batching too, but I know that complicates the current example code & bucketing executor, so not really a requirement! Thanks again for getting an example of doing inference in the c++ api. This is an automated message from the Apache Git Service. To respond to the message, please log on 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] stephenrawls commented on a change in pull request #13680: [MXNET-1121] Example to demonstrate the inference workflow using RNN
stephenrawls commented on a change in pull request #13680: [MXNET-1121] Example
to demonstrate the inference workflow using RNN
URL: https://github.com/apache/incubator-mxnet/pull/13680#discussion_r253278924
##
File path: cpp-package/example/inference/sentiment_analysis_rnn.cpp
##
@@ -0,0 +1,397 @@
+/*
+ * 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.
+ */
+
+/*
+ * This example demonstrates sentiment prediction workflow with pre-trained
RNN model using MXNet C++ API.
+ * The example performs following tasks.
+ * 1. Load the pre-trained RNN model,
+ * 2. Load the dictionary file that contains word to index mapping.
+ * 3. Convert the input string to vector of indices and padded to match the
input data length.
+ * 4. Run the forward pass and predict the output string.
+ * The example uses a pre-trained RNN model that is trained with the IMDB
dataset.
+ */
+
+#include
+#include
+#include
+#include
+#include
+#include
+#include
+#include
+#include "mxnet-cpp/MxNetCpp.h"
+
+using namespace mxnet::cpp;
+
+static const int DEFAULT_NUM_WORDS = 5;
+static const char DEFAULT_S3_URL[] =
"https://s3.amazonaws.com/mxnet-cpp/RNN_model/";;
+
+/*
+ * class Predictor
+ *
+ * This class encapsulates the functionality to load the model, process input
image and run the forward pass.
+ */
+
+class Predictor {
+ public:
+Predictor() {}
+Predictor(const std::string& model_json,
+ const std::string& model_params,
+ const std::string& input_dictionary,
+ bool use_gpu = false,
+ int num_words = DEFAULT_NUM_WORDS);
+float PredictSentiment(const std::string &input_sequence);
+~Predictor();
+
+ private:
+void LoadModel(const std::string& model_json_file);
+void LoadParameters(const std::string& model_parameters_file);
+void LoadDictionary(const std::string &input_dictionary);
+inline bool FileExists(const std::string& name) {
+struct stat buffer;
+return (stat(name.c_str(), &buffer) == 0);
+}
+int ConverToIndexVector(const std::string& input,
+ std::vector *input_vector);
+int GetIndexForOutputSymbolName(const std::string& output_symbol_name);
+float GetIndexForWord(const std::string& word);
+std::map args_map;
+std::map aux_map;
+std::map wordToIndex;
+Symbol net;
+Executor *executor;
+Context global_ctx = Context::cpu();
+int num_words;
+};
+
+
+/*
+ * The constructor takes the following parameters as input:
+ * 1. model_json: The RNN model in json formatted file.
+ * 2. model_params: File containing model parameters
+ * 3. input_dictionary: File containing the word and associated index.
+ * 4. num_words: Number of words which will be used to predict the sentiment.
+ *
+ * The constructor:
+ * 1. Loads the model and parameter files.
+ * 2. Loads the dictionary file to create index to word and word to index
maps.
+ * 3. Invokes the SimpleBind to bind the input argument to the model and
create an executor.
+ *
+ * The SimpleBind is expected to be invoked only once.
+ */
+Predictor::Predictor(const std::string& model_json,
+ const std::string& model_params,
+ const std::string& input_dictionary,
+ bool use_gpu,
+ int num_words):num_words(num_words) {
+ if (use_gpu) {
+global_ctx = Context::gpu();
+ }
+
+ /*
+ * Load the dictionary file that contains the word and its index.
+ * The function creates word to index and index to word map. The maps are
used to create index
+ * vector for the input sentence.
+ */
+ LoadDictionary(input_dictionary);
+
+ // Load the model
+ LoadModel(model_json);
+
+ // Load the model parameters.
+ LoadParameters(model_params);
+
+ args_map["data0"] = NDArray(Shape(num_words, 1), global_ctx, false);
+ args_map["data1"] = NDArray(Shape(1), global_ctx, false);
+
+ executor = net.SimpleBind(global_ctx, args_map, std::map(),
+ std::map(), aux_map);
+}
+
+
+/*
+ * The following function loads the model from json file.
+ */
+void Predictor::LoadModel(const std::string& model_json_file) {
+ if (!FileExists(model_json_file)) {
[GitHub] stephenrawls commented on a change in pull request #13680: [MXNET-1121] Example to demonstrate the inference workflow using RNN
stephenrawls commented on a change in pull request #13680: [MXNET-1121] Example
to demonstrate the inference workflow using RNN
URL: https://github.com/apache/incubator-mxnet/pull/13680#discussion_r253278383
##
File path: cpp-package/example/inference/sentiment_analysis_rnn.cpp
##
@@ -0,0 +1,464 @@
+/*
+ * 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.
+ */
+
+/*
+ * This example demonstrates sentiment prediction workflow with pre-trained
RNN model using MXNet C++ API.
+ * The example performs following tasks.
+ * 1. Load the pre-trained RNN model,
+ * 2. Load the dictionary file that contains word to index mapping.
+ * 3. Create executors for pre-determined input lengths.
+ * 4. Convert each line in the input to the vector of indices.
+ * 5. Predictor finds the right executor for each line.
+ * 4. Run the forward pass for each line and predicts the sentiment scores.
+ * The example uses a pre-trained RNN model that is trained with the IMDB
dataset.
+ */
+
+#include
+#include
+#include
+#include
+#include
+#include
+#include
+#include
+#include "mxnet-cpp/MxNetCpp.h"
+
+using namespace mxnet::cpp;
+
+static const int DEFAULT_BUCKET_KEYS[] = {5, 10, 15, 20, 25, 30};
+static const char DEFAULT_S3_URL[] =
"https://s3.amazonaws.com/mxnet-cpp/RNN_model/";;
+
+/*
+ * class Predictor
+ *
+ * This class encapsulates the functionality to load the model, process input
image and run the forward pass.
+ */
+
+class Predictor {
+ public:
+Predictor() {}
+Predictor(const std::string& model_json,
+ const std::string& model_params,
+ const std::string& input_dictionary,
+ const std::vector& bucket_keys,
+ bool use_gpu = false);
+float PredictSentiment(const std::string &input_review);
+~Predictor();
+
+ private:
+void LoadModel(const std::string& model_json_file);
+void LoadParameters(const std::string& model_parameters_file);
+void LoadDictionary(const std::string &input_dictionary);
+inline bool FileExists(const std::string& name) {
+struct stat buffer;
+return (stat(name.c_str(), &buffer) == 0);
+}
+float PredictSentimentForOneLine(const std::string &input_line);
+int ConvertToIndexVector(const std::string& input,
+ std::vector *input_vector);
+int GetIndexForOutputSymbolName(const std::string& output_symbol_name);
+float GetIndexForWord(const std::string& word);
+int GetClosestBucketKey(int num_words);
+std::map args_map;
+std::map aux_map;
+std::map wordToIndex;
+Symbol net;
+std::map executor_buckets;
+Context global_ctx = Context::cpu();
+};
+
+
+/*
+ * The constructor takes the following parameters as input:
+ * 1. model_json: The RNN model in json formatted file.
+ * 2. model_params: File containing model parameters
+ * 3. input_dictionary: File containing the word and associated index.
+ * 4. num_words: Number of words which will be used to predict the sentiment.
+ *
+ * The constructor:
+ * 1. Loads the model and parameter files.
+ * 2. Loads the dictionary file to create index to word and word to index
maps.
+ * 3. For each bucket key in the input vector of bucket keys, it invokes the
SimpleBind to
+ * create the executor. The bucket key determines the length of input data
required
+ * for that executor.
+ * 4. Creates a map of bucket key to corresponding executor.
+ * 5. The model is loaded only once. The executors share the memory for the
parameters.
+ */
+Predictor::Predictor(const std::string& model_json,
+ const std::string& model_params,
+ const std::string& input_dictionary,
+ const std::vector& bucket_keys,
+ bool use_gpu) {
+ if (use_gpu) {
+global_ctx = Context::gpu();
+ }
+
+ /*
+ * Load the dictionary file that contains the word and its index.
+ * The function creates word to index and index to word map. The maps are
used to create index
+ * vector for the input sentence.
+ */
+ LoadDictionary(input_dictionary);
+
+ // Load the model
+ LoadModel(model_json);
+
+ // Load the model parameters.
+ LoadParameters(model_params);
+
+
[incubator-mxnet-site] branch asf-site updated: Bump the publish timestamp.
This is an automated email from the ASF dual-hosted git repository. zhasheng 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 67dc46d Bump the publish timestamp. 67dc46d is described below commit 67dc46d713b1486507aaff61fc991e10a3aa8843 Author: mxnet-ci AuthorDate: Sat Feb 2 20:40:43 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..1789ea1 --- /dev/null +++ b/date.txt @@ -0,0 +1 @@ +Sat Feb 2 20:40:43 UTC 2019
[incubator-mxnet-site] branch asf-site updated: Bump the publish timestamp.
This is an automated email from the ASF dual-hosted git repository. zhasheng 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 ba27f54 Bump the publish timestamp. ba27f54 is described below commit ba27f54dd3072f02ce834e53dc108053272dea97 Author: mxnet-ci AuthorDate: Sat Feb 2 19:07:14 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..118832a --- /dev/null +++ b/date.txt @@ -0,0 +1 @@ +Sat Feb 2 19:07:14 UTC 2019
[GitHub] szha commented on issue #14053: in-place reshape ops
szha commented on issue #14053: in-place reshape ops URL: https://github.com/apache/incubator-mxnet/pull/14053#issuecomment-459990251 @ZhennanQin I didn't test that specifically based on the assumption that the backend implementation should not cause existing APIs such as `NDArray.reshape` or `NDArray.shape` to break. This is an automated message from the Apache Git Service. To respond to the message, please log on 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] renganxu edited a comment on issue #14047: mxnet.base.MXNetError: Cannot find argument 'cudnn_algo_verbose'
renganxu edited a comment on issue #14047: mxnet.base.MXNetError: Cannot find argument 'cudnn_algo_verbose' URL: https://github.com/apache/incubator-mxnet/issues/14047#issuecomment-459983879 @ptrendx I know NGC container works, but I just want to run without container. Now I figured out NGC MXNet changed the Convolution operator by adding more parameters: cudnn_algo_verbose, cudnn_algo_fwd, cudnn_algo_bwd_data, cudnn_algo_bwd_filter, and cudnn_tensor_core_only. But they are not available in MXNet repo. Do you know whether the MXNet repo has any plan to integrate these changes made by Nvidia? Thanks. This is an automated message from the Apache Git Service. To respond to the message, please log on 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] renganxu commented on issue #14047: mxnet.base.MXNetError: Cannot find argument 'cudnn_algo_verbose'
renganxu commented on issue #14047: mxnet.base.MXNetError: Cannot find argument 'cudnn_algo_verbose' URL: https://github.com/apache/incubator-mxnet/issues/14047#issuecomment-459983879 @ptrendx I know NGC container works, but I just want to run without container. Now I figured out NGC MXNet changed the Convolution operator by adding more parameters: cudnn_algo_verbose, cudnn_algo_fwd, cudnn_algo_bwd_data, cudnn_algo_bwd_filter, and cudnn_tensor_core_only. Do you know whether the MXNet repo has any plan to integrate these changes made by Nvidia? Thanks. This is an automated message from the Apache Git Service. To respond to the message, please log on 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] larroy closed pull request #13834: Reduce maven verbosity which is filling up the logs of the builds
larroy closed pull request #13834: Reduce maven verbosity which is filling up the logs of the builds URL: https://github.com/apache/incubator-mxnet/pull/13834 This is an automated message from the Apache Git Service. To respond to the message, please log on 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] mxnet-label-bot commented on issue #14059: ModuleNotFoundError: No module named 'mxnet'
mxnet-label-bot commented on issue #14059: ModuleNotFoundError: No module named 'mxnet' URL: https://github.com/apache/incubator-mxnet/issues/14059#issuecomment-459979035 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 labels: Installation This is an automated message from the Apache Git Service. To respond to the message, please log on 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] mahmoodn opened a new issue #14059: ModuleNotFoundError: No module named 'mxnet'
mahmoodn opened a new issue #14059: ModuleNotFoundError: No module named 'mxnet' URL: https://github.com/apache/incubator-mxnet/issues/14059 Hi, Although I have built mxnet with `make -j8 USE_OPENCV=1 USE_BLAS=openblas USE_CUDA=1 USE_CUDA_PATH=/usr/local/cuda USE_CUDNN=1`, I am not able to run an example $ python main.py --configfile default.cfg Traceback (most recent call last): File "main.py", line 23, in from config_util import parse_args, parse_contexts, generate_file_path File "/home/mahmood/mx/mxnet/example/speech_recognition/config_util.py", line 23, in import mxnet as mx ModuleNotFoundError: No module named 'mxnet' I think I missed some steps. As stated `[here](https://mxnet.incubator.apache.org/versions/master/install/build_from_source.html)`, there a Ninja thing! Is that mandatory? This is an automated message from the Apache Git Service. To respond to the message, please log on 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] gigasquid commented on a change in pull request #13993: [Clojure] Add resource scope to clojure package
gigasquid commented on a change in pull request #13993: [Clojure] Add resource
scope to clojure package
URL: https://github.com/apache/incubator-mxnet/pull/13993#discussion_r253270361
##
File path:
contrib/clojure-package/test/org/apache/clojure_mxnet/resource_scope_test.clj
##
@@ -0,0 +1,149 @@
+;;
+;; 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.
+;;
+
+(ns org.apache.clojure-mxnet.resource-scope-test
+ (:require [org.apache.clojure-mxnet.ndarray :as ndarray]
+[org.apache.clojure-mxnet.symbol :as sym]
+[org.apache.clojure-mxnet.resource-scope :as resource-scope]
+[clojure.test :refer :all]))
+
+
+(deftest test-resource-scope-with-ndarray
+ (let [native-resources (atom {})
+x (ndarray/ones [2 2])
+return-val (resource-scope/using
+(let [temp-x (ndarray/ones [3 1])
+ temp-y (ndarray/ones [3 1])]
+ (swap! native-resources assoc :temp-x temp-x)
+ (swap! native-resources assoc :temp-y temp-y)
+ (ndarray/+ temp-x 1)))]
+(is (true? (ndarray/is-disposed (:temp-x @native-resources
+(is (true? (ndarray/is-disposed (:temp-y @native-resources
+(is (false? (ndarray/is-disposed return-val)))
+(is (false? (ndarray/is-disposed x)))
+(is (= [2.0 2.0 2.0] (ndarray/->vec return-val)
+
+(deftest test-nested-resource-scope-with-ndarray
+ (let [native-resources (atom {})
+x (ndarray/ones [2 2])
+return-val (resource-scope/using
+(let [temp-x (ndarray/ones [3 1])]
+ (swap! native-resources assoc :temp-x temp-x)
+ (resource-scope/using
+ (let [temp-y (ndarray/ones [3 1])]
+(swap! native-resources assoc :temp-y temp-y)]
+(is (true? (ndarray/is-disposed (:temp-y @native-resources
+(is (true? (ndarray/is-disposed (:temp-x @native-resources
+(is (false? (ndarray/is-disposed x)
+
+(deftest test-resource-scope-with-sym
+ (let [native-resources (atom {})
+x (sym/ones [2 2])
+return-val (resource-scope/using
+(let [temp-x (sym/ones [3 1])
+ temp-y (sym/ones [3 1])]
+ (swap! native-resources assoc :temp-x temp-x)
+ (swap! native-resources assoc :temp-y temp-y)
+ (sym/+ temp-x 1)))]
+(is (true? (sym/is-disposed (:temp-x @native-resources
+(is (true? (sym/is-disposed (:temp-y @native-resources
+(is (false? (sym/is-disposed return-val)))
+(is (false? (sym/is-disposed x)
+
+(deftest test-nested-resource-scope-with-ndarray
+ (let [native-resources (atom {})
+x (ndarray/ones [2 2])
+return-val (resource-scope/using
+(let [temp-x (ndarray/ones [3 1])]
+ (swap! native-resources assoc :temp-x temp-x)
+ (resource-scope/using
+ (let [temp-y (ndarray/ones [3 1])]
+(swap! native-resources assoc :temp-y temp-y)]
+(is (true? (ndarray/is-disposed (:temp-y @native-resources
+(is (true? (ndarray/is-disposed (:temp-x @native-resources
+(is (false? (ndarray/is-disposed x)
+
+(deftest test-nested-resource-scope-with-sym
+ (let [native-resources (atom {})
+x (sym/ones [2 2])
+return-val (resource-scope/using
+(let [temp-x (sym/ones [3 1])]
+ (swap! native-resources assoc :temp-x temp-x)
+ (resource-scope/using
+ (let [temp-y (sym/ones [3 1])]
+(swap! native-resources assoc :temp-y temp-y)]
+(is (true? (sym/is-disposed (:temp-y @native-resources
+(is (true? (sym/is-disposed (:temp-x @native-resources
+(is (false? (sym/is-disposed x)
+
+;;; Note that if first is returned the rest of the collection ndarrays will
+;;; NOT be disposed
Review comment:
oh yes - the curse of the leftover misleading comments :)
This is an automated message fr
[GitHub] kedarbellare commented on issue #13993: [Clojure] Add resource scope to clojure package
kedarbellare commented on issue #13993: [Clojure] Add resource scope to clojure package URL: https://github.com/apache/incubator-mxnet/pull/13993#issuecomment-459977844 can't wait to use this 😃 This is an automated message from the Apache Git Service. To respond to the message, please log on 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] kedarbellare commented on a change in pull request #13993: [Clojure] Add resource scope to clojure package
kedarbellare commented on a change in pull request #13993: [Clojure] Add
resource scope to clojure package
URL: https://github.com/apache/incubator-mxnet/pull/13993#discussion_r253270291
##
File path:
contrib/clojure-package/test/org/apache/clojure_mxnet/resource_scope_test.clj
##
@@ -0,0 +1,149 @@
+;;
+;; 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.
+;;
+
+(ns org.apache.clojure-mxnet.resource-scope-test
+ (:require [org.apache.clojure-mxnet.ndarray :as ndarray]
+[org.apache.clojure-mxnet.symbol :as sym]
+[org.apache.clojure-mxnet.resource-scope :as resource-scope]
+[clojure.test :refer :all]))
+
+
+(deftest test-resource-scope-with-ndarray
+ (let [native-resources (atom {})
+x (ndarray/ones [2 2])
+return-val (resource-scope/using
+(let [temp-x (ndarray/ones [3 1])
+ temp-y (ndarray/ones [3 1])]
+ (swap! native-resources assoc :temp-x temp-x)
+ (swap! native-resources assoc :temp-y temp-y)
+ (ndarray/+ temp-x 1)))]
+(is (true? (ndarray/is-disposed (:temp-x @native-resources
+(is (true? (ndarray/is-disposed (:temp-y @native-resources
+(is (false? (ndarray/is-disposed return-val)))
+(is (false? (ndarray/is-disposed x)))
+(is (= [2.0 2.0 2.0] (ndarray/->vec return-val)
+
+(deftest test-nested-resource-scope-with-ndarray
+ (let [native-resources (atom {})
+x (ndarray/ones [2 2])
+return-val (resource-scope/using
+(let [temp-x (ndarray/ones [3 1])]
+ (swap! native-resources assoc :temp-x temp-x)
+ (resource-scope/using
+ (let [temp-y (ndarray/ones [3 1])]
+(swap! native-resources assoc :temp-y temp-y)]
+(is (true? (ndarray/is-disposed (:temp-y @native-resources
+(is (true? (ndarray/is-disposed (:temp-x @native-resources
+(is (false? (ndarray/is-disposed x)
+
+(deftest test-resource-scope-with-sym
+ (let [native-resources (atom {})
+x (sym/ones [2 2])
+return-val (resource-scope/using
+(let [temp-x (sym/ones [3 1])
+ temp-y (sym/ones [3 1])]
+ (swap! native-resources assoc :temp-x temp-x)
+ (swap! native-resources assoc :temp-y temp-y)
+ (sym/+ temp-x 1)))]
+(is (true? (sym/is-disposed (:temp-x @native-resources
+(is (true? (sym/is-disposed (:temp-y @native-resources
+(is (false? (sym/is-disposed return-val)))
+(is (false? (sym/is-disposed x)
+
+(deftest test-nested-resource-scope-with-ndarray
+ (let [native-resources (atom {})
+x (ndarray/ones [2 2])
+return-val (resource-scope/using
+(let [temp-x (ndarray/ones [3 1])]
+ (swap! native-resources assoc :temp-x temp-x)
+ (resource-scope/using
+ (let [temp-y (ndarray/ones [3 1])]
+(swap! native-resources assoc :temp-y temp-y)]
+(is (true? (ndarray/is-disposed (:temp-y @native-resources
+(is (true? (ndarray/is-disposed (:temp-x @native-resources
+(is (false? (ndarray/is-disposed x)
+
+(deftest test-nested-resource-scope-with-sym
+ (let [native-resources (atom {})
+x (sym/ones [2 2])
+return-val (resource-scope/using
+(let [temp-x (sym/ones [3 1])]
+ (swap! native-resources assoc :temp-x temp-x)
+ (resource-scope/using
+ (let [temp-y (sym/ones [3 1])]
+(swap! native-resources assoc :temp-y temp-y)]
+(is (true? (sym/is-disposed (:temp-y @native-resources
+(is (true? (sym/is-disposed (:temp-x @native-resources
+(is (false? (sym/is-disposed x)
+
+;;; Note that if first is returned the rest of the collection ndarrays will
+;;; NOT be disposed
Review comment:
update comment?
This is an automated message from the Apache Git Service.
To respond t
[GitHub] gigasquid commented on a change in pull request #13993: [Clojure] Add resource scope to clojure package
gigasquid commented on a change in pull request #13993: [Clojure] Add resource
scope to clojure package
URL: https://github.com/apache/incubator-mxnet/pull/13993#discussion_r253269849
##
File path:
contrib/clojure-package/examples/imclassification/src/imclassification/train_mnist.clj
##
@@ -96,18 +75,33 @@
(do
(println "Starting Training of MNIST ")
(println "Running with context devices of" devs)
- (let [_mod (m/module (get-symbol) {:contexts devs})]
-(m/fit _mod {:train-data train-data
-:eval-data test-data
+ (resource-scope/with-let [_mod (m/module (get-symbol) {:contexts devs})]
+(-> _mod
+(m/fit {:train-data (mx-io/mnist-iter {:image (str data-dir
"train-images-idx3-ubyte")
Review comment:
It will work if they are `defs` outside, but you will still get the
undisposed ndarray warning. I refactored to move it out to functions which does
work and looks better too :)
This is an automated message from the Apache Git Service.
To respond to the message, please log on 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] gigasquid commented on a change in pull request #13993: [Clojure] Add resource scope to clojure package
gigasquid commented on a change in pull request #13993: [Clojure] Add resource
scope to clojure package
URL: https://github.com/apache/incubator-mxnet/pull/13993#discussion_r253269610
##
File path:
contrib/clojure-package/test/org/apache/clojure_mxnet/resource_scope_test.clj
##
@@ -0,0 +1,151 @@
+;;
+;; 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.
+;;
+
+(ns org.apache.clojure-mxnet.resource-scope-test
+ (:require [org.apache.clojure-mxnet.ndarray :as ndarray]
+[org.apache.clojure-mxnet.symbol :as sym]
+[org.apache.clojure-mxnet.resource-scope :as resource-scope]
+[clojure.test :refer :all]))
+
+
+(deftest test-resource-scope-with-ndarray
+ (let [native-resources (atom {})
+x (ndarray/ones [2 2])
+return-val (resource-scope/using
+(let [temp-x (ndarray/ones [3 1])
+ temp-y (ndarray/ones [3 1])]
+ (swap! native-resources assoc :temp-x temp-x)
+ (swap! native-resources assoc :temp-y temp-y)
+ (ndarray/+ temp-x 1)))]
+(is (true? (ndarray/is-disposed (:temp-x @native-resources
+(is (true? (ndarray/is-disposed (:temp-y @native-resources
+(is (false? (ndarray/is-disposed return-val)))
+(is (false? (ndarray/is-disposed x)))
+(is (= [2.0 2.0 2.0] (ndarray/->vec return-val)
+
+(deftest test-nested-resource-scope-with-ndarray
+ (let [native-resources (atom {})
+x (ndarray/ones [2 2])
+return-val (resource-scope/using
+(let [temp-x (ndarray/ones [3 1])]
+ (swap! native-resources assoc :temp-x temp-x)
+ (resource-scope/using
+ (let [temp-y (ndarray/ones [3 1])]
+(swap! native-resources assoc :temp-y temp-y)]
+(is (true? (ndarray/is-disposed (:temp-y @native-resources
+(is (true? (ndarray/is-disposed (:temp-x @native-resources
+(is (false? (ndarray/is-disposed x)
+
+(deftest test-resource-scope-with-sym
+ (let [native-resources (atom {})
+x (sym/ones [2 2])
+return-val (resource-scope/using
+(let [temp-x (sym/ones [3 1])
+ temp-y (sym/ones [3 1])]
+ (swap! native-resources assoc :temp-x temp-x)
+ (swap! native-resources assoc :temp-y temp-y)
+ (sym/+ temp-x 1)))]
+(is (true? (sym/is-disposed (:temp-x @native-resources
+(is (true? (sym/is-disposed (:temp-y @native-resources
+(is (false? (sym/is-disposed return-val)))
+(is (false? (sym/is-disposed x)
+
+(deftest test-nested-resource-scope-with-ndarray
+ (let [native-resources (atom {})
+x (ndarray/ones [2 2])
+return-val (resource-scope/using
+(let [temp-x (ndarray/ones [3 1])]
+ (swap! native-resources assoc :temp-x temp-x)
+ (resource-scope/using
+ (let [temp-y (ndarray/ones [3 1])]
+(swap! native-resources assoc :temp-y temp-y)]
+(is (true? (ndarray/is-disposed (:temp-y @native-resources
+(is (true? (ndarray/is-disposed (:temp-x @native-resources
+(is (false? (ndarray/is-disposed x)
+
+(deftest test-nested-resource-scope-with-sym
+ (let [native-resources (atom {})
+x (sym/ones [2 2])
+return-val (resource-scope/using
+(let [temp-x (sym/ones [3 1])]
+ (swap! native-resources assoc :temp-x temp-x)
+ (resource-scope/using
+ (let [temp-y (sym/ones [3 1])]
+(swap! native-resources assoc :temp-y temp-y)]
+(is (true? (sym/is-disposed (:temp-y @native-resources
+(is (true? (sym/is-disposed (:temp-x @native-resources
+(is (false? (sym/is-disposed x)
+
+;;; Note that if first is returned the rest of the collection ndarrays will
+;;; NOT be disposed
+(deftest test-list-creation-with-returning-first
+ (let [native-resources (atom {})
+return-val (resource-scope/using
+(let [temp-ndarrays (mapv (c
[GitHub] mahmoodn closed issue #14002: error: ‘__cpuid’ was not declared in this scope
mahmoodn closed issue #14002: error: ‘__cpuid’ was not declared in this scope URL: https://github.com/apache/incubator-mxnet/issues/14002 This is an automated message from the Apache Git Service. To respond to the message, please log on 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] mahmoodn commented on issue #14002: error: ‘__cpuid’ was not declared in this scope
mahmoodn commented on issue #14002: error: ‘__cpuid’ was not declared in this scope URL: https://github.com/apache/incubator-mxnet/issues/14002#issuecomment-459975490 It seems that I was able to compile the latest git version. I will close the issue. The last thing in the build output is ``` g++ -DMSHADOW_FORCE_STREAM -Wall -Wsign-compare -O3 -DNDEBUG=1 -I/home/mahmood/mx/mxnet/3rdparty/mshadow/ -I/home/mahmood/mx/mxnet/3rdparty/dmlc-core/include -fPIC -I/home/mahmood/mx/mxnet/3rdparty/tvm/nnvm/include -I/home/mahmood/mx/mxnet/3rdparty/dlpack/include -I/home/mahmood/mx/mxnet/3rdparty/tvm/include -Iinclude -funroll-loops -Wno-unused-parameter -Wno-unknown-pragmas -Wno-unused-local-typedefs -msse3 -mf16c -I/usr/local/cuda/include -DMSHADOW_USE_CBLAS=1 -DMSHADOW_USE_MKL=0 -I/home/mahmood/mx/mxnet/3rdparty/mkldnn/build/install/include -DMSHADOW_RABIT_PS=0 -DMSHADOW_DIST_PS=0 -DMSHADOW_USE_PASCAL=0 -DMXNET_USE_MKLDNN=1 -DUSE_MKL=1 -I/home/mahmood/mx/mxnet/src/operator/nn/mkldnn/ -I/home/mahmood/mx/mxnet/3rdparty/mkldnn/build/install/include -DMXNET_USE_OPENCV=1 -I/usr/include/opencv -fopenmp -DMXNET_USE_OPERATOR_TUNING=1 -DMSHADOW_USE_CUDNN=1 -I/home/mahmood/mx/mxnet/3rdparty/cub -DMXNET_ENABLE_CUDA_RTC=1 -DMXNET_USE_NCCL=0 -DMXNET_USE_LIBJPEG_TURBO=0 -shared -o lib/libmxnet.so build/src/operator/quantization/mkldnn/mkldnn_quantized_conv.o build/src/operator/quantization/mkldnn/mkldnn_quantized_pooling.o build/src/operator/quantization/mkldnn/mkldnn_quantized_concat.o build/src/operator/subgraph/mkldnn/mkldnn_conv_property.o build/src/operator/subgraph/mkldnn/mkldnn_conv_post_quantize_property.o build/src/operator/subgraph/mkldnn/mkldnn_conv.o build/src/operator/nn/mkldnn/mkldnn_convolution.o build/src/operator/nn/mkldnn/mkldnn_concat.o build/src/operator/nn/mkldnn/mkldnn_base.o build/src/operator/nn/mkldnn/mkldnn_slice.o build/src/operator/nn/mkldnn/mkldnn_act.o build/src/operator/nn/mkldnn/mkldnn_softmax.o build/src/operator/nn/mkldnn/mkldnn_deconvolution.o build/src/operator/nn/mkldnn/mkldnn_copy.o build/src/operator/nn/mkldnn/mkldnn_fully_connected.o build/src/operator/nn/mkldnn/mkldnn_pooling.o build/src/operator/nn/mkldnn/mkldnn_sum.o build/src/operator/nn/cudnn/cudnn_algoreg.o build/src/operator/nn/cudnn/cudnn_batch_norm.o build/src/operator/tensor/elemwise_binary_broadcast_op_basic.o build/src/operator/tensor/elemwise_binary_op_logic.o build/src/operator/tensor/square_sum.o build/src/operator/tensor/matrix_op.o build/src/operator/tensor/init_op.o build/src/operator/tensor/elemwise_sum.o build/src/operator/tensor/la_op.o build/src/operator/tensor/histogram.o build/src/operator/tensor/broadcast_reduce_op_index.o build/src/operator/tensor/dot.o build/src/operator/tensor/elemwise_scatter_op.o build/src/operator/tensor/elemwise_unary_op_basic.o build/src/operator/tensor/elemwise_binary_broadcast_op_extended.o build/src/operator/tensor/ravel.o build/src/operator/tensor/broadcast_reduce_op_value.o build/src/operator/tensor/control_flow_op.o build/src/operator/tensor/elemwise_binary_op_basic.o build/src/operator/tensor/elemwise_binary_op_extended.o build/src/operator/tensor/indexing_op.o build/src/operator/tensor/elemwise_binary_broadcast_op_logic.o build/src/operator/tensor/diag_op.o build/src/operator/tensor/ordering_op.o build/src/operator/tensor/sparse_retain.o build/src/operator/tensor/elemwise_binary_scalar_op_extended.o build/src/operator/tensor/elemwise_binary_scalar_op_basic.o build/src/operator/tensor/elemwise_binary_scalar_op_logic.o build/src/operator/tensor/cast_storage.o build/src/operator/tensor/elemwise_binary_op.o build/src/operator/tensor/elemwise_unary_op_trig.o build/src/operator/contrib/tensorrt.o build/src/operator/contrib/multibox_target.o build/src/operator/contrib/sync_batch_norm.o build/src/operator/contrib/count_sketch.o build/src/operator/contrib/roi_align.o build/src/operator/contrib/bilinear_resize.o build/src/operator/contrib/nnz.o build/src/operator/contrib/multibox_detection.o build/src/operator/contrib/nnvm_to_onnx.o build/src/operator/contrib/deformable_psroi_pooling.o build/src/operator/contrib/dgl_graph.o build/src/operator/contrib/fft.o build/src/operator/contrib/multibox_prior.o build/src/operator/contrib/gradient_multiplier_op.o build/src/operator/contrib/adamw.o build/src/operator/contrib/transformer.o build/src/operator/contrib/krprod.o build/src/operator/contrib/multi_proposal.o build/src/operator/contrib/index_copy.o build/src/operator/contrib/optimizer_op.o build/src/operator/contrib/bounding_box.o build/src/operator/contrib/proposal.o build/src/operator/contrib/boolean_mask.o build/src/operator/contrib/psroi_pooling.o build/src/operator/contrib/quadratic_op.o build/src/operator/contrib/deformable_convolution.o build/src/operator/contrib/ifft.o build/src/operator/contrib/adaptive_avg_pooling.o build/src/operator/random/sample_multinomial_op.o build/src/operator/ra
[GitHub] kedarbellare commented on a change in pull request #13993: [Clojure] Add resource scope to clojure package
kedarbellare commented on a change in pull request #13993: [Clojure] Add
resource scope to clojure package
URL: https://github.com/apache/incubator-mxnet/pull/13993#discussion_r253268961
##
File path:
contrib/clojure-package/examples/imclassification/src/imclassification/train_mnist.clj
##
@@ -96,18 +75,33 @@
(do
(println "Starting Training of MNIST ")
(println "Running with context devices of" devs)
- (let [_mod (m/module (get-symbol) {:contexts devs})]
-(m/fit _mod {:train-data train-data
-:eval-data test-data
+ (resource-scope/with-let [_mod (m/module (get-symbol) {:contexts devs})]
+(-> _mod
+(m/fit {:train-data (mx-io/mnist-iter {:image (str data-dir
"train-images-idx3-ubyte")
Review comment:
just for my understanding, does `resource-scope` not work when the
`train-data` and `eval-data` are defined outside the `with-let` block?
This is an automated message from the Apache Git Service.
To respond to the message, please log on 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] kedarbellare commented on a change in pull request #13993: [Clojure] Add resource scope to clojure package
kedarbellare commented on a change in pull request #13993: [Clojure] Add
resource scope to clojure package
URL: https://github.com/apache/incubator-mxnet/pull/13993#discussion_r253268889
##
File path:
contrib/clojure-package/test/org/apache/clojure_mxnet/resource_scope_test.clj
##
@@ -0,0 +1,151 @@
+;;
+;; 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.
+;;
+
+(ns org.apache.clojure-mxnet.resource-scope-test
+ (:require [org.apache.clojure-mxnet.ndarray :as ndarray]
+[org.apache.clojure-mxnet.symbol :as sym]
+[org.apache.clojure-mxnet.resource-scope :as resource-scope]
+[clojure.test :refer :all]))
+
+
+(deftest test-resource-scope-with-ndarray
+ (let [native-resources (atom {})
+x (ndarray/ones [2 2])
+return-val (resource-scope/using
+(let [temp-x (ndarray/ones [3 1])
+ temp-y (ndarray/ones [3 1])]
+ (swap! native-resources assoc :temp-x temp-x)
+ (swap! native-resources assoc :temp-y temp-y)
+ (ndarray/+ temp-x 1)))]
+(is (true? (ndarray/is-disposed (:temp-x @native-resources
+(is (true? (ndarray/is-disposed (:temp-y @native-resources
+(is (false? (ndarray/is-disposed return-val)))
+(is (false? (ndarray/is-disposed x)))
+(is (= [2.0 2.0 2.0] (ndarray/->vec return-val)
+
+(deftest test-nested-resource-scope-with-ndarray
+ (let [native-resources (atom {})
+x (ndarray/ones [2 2])
+return-val (resource-scope/using
+(let [temp-x (ndarray/ones [3 1])]
+ (swap! native-resources assoc :temp-x temp-x)
+ (resource-scope/using
+ (let [temp-y (ndarray/ones [3 1])]
+(swap! native-resources assoc :temp-y temp-y)]
+(is (true? (ndarray/is-disposed (:temp-y @native-resources
+(is (true? (ndarray/is-disposed (:temp-x @native-resources
+(is (false? (ndarray/is-disposed x)
+
+(deftest test-resource-scope-with-sym
+ (let [native-resources (atom {})
+x (sym/ones [2 2])
+return-val (resource-scope/using
+(let [temp-x (sym/ones [3 1])
+ temp-y (sym/ones [3 1])]
+ (swap! native-resources assoc :temp-x temp-x)
+ (swap! native-resources assoc :temp-y temp-y)
+ (sym/+ temp-x 1)))]
+(is (true? (sym/is-disposed (:temp-x @native-resources
+(is (true? (sym/is-disposed (:temp-y @native-resources
+(is (false? (sym/is-disposed return-val)))
+(is (false? (sym/is-disposed x)
+
+(deftest test-nested-resource-scope-with-ndarray
+ (let [native-resources (atom {})
+x (ndarray/ones [2 2])
+return-val (resource-scope/using
+(let [temp-x (ndarray/ones [3 1])]
+ (swap! native-resources assoc :temp-x temp-x)
+ (resource-scope/using
+ (let [temp-y (ndarray/ones [3 1])]
+(swap! native-resources assoc :temp-y temp-y)]
+(is (true? (ndarray/is-disposed (:temp-y @native-resources
+(is (true? (ndarray/is-disposed (:temp-x @native-resources
+(is (false? (ndarray/is-disposed x)
+
+(deftest test-nested-resource-scope-with-sym
+ (let [native-resources (atom {})
+x (sym/ones [2 2])
+return-val (resource-scope/using
+(let [temp-x (sym/ones [3 1])]
+ (swap! native-resources assoc :temp-x temp-x)
+ (resource-scope/using
+ (let [temp-y (sym/ones [3 1])]
+(swap! native-resources assoc :temp-y temp-y)]
+(is (true? (sym/is-disposed (:temp-y @native-resources
+(is (true? (sym/is-disposed (:temp-x @native-resources
+(is (false? (sym/is-disposed x)
+
+;;; Note that if first is returned the rest of the collection ndarrays will
+;;; NOT be disposed
+(deftest test-list-creation-with-returning-first
+ (let [native-resources (atom {})
+return-val (resource-scope/using
+(let [temp-ndarrays (mapv
[GitHub] TaoLv commented on issue #13749: Add NHWC layout support to Pooling (cpu, gpu cuda, gpu cuDNN)
TaoLv commented on issue #13749: Add NHWC layout support to Pooling (cpu, gpu cuda, gpu cuDNN) URL: https://github.com/apache/incubator-mxnet/pull/13749#issuecomment-459970152 Maybe it's out of the scope of this PR, but I would suggest to remove layout parameters from all operator APIs in the next major release. It's framework that should take care of the performance of model. We cannot ask users to complicate their models to get some performance benefit from one specific hardware or backend. The change may be nightmare if they want to change back to other hardwares or backends. This is an automated message from the Apache Git Service. To respond to the message, please log on 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] TaoLv commented on a change in pull request #13749: Add NHWC layout support to Pooling (cpu, gpu cuda, gpu cuDNN)
TaoLv commented on a change in pull request #13749: Add NHWC layout support to
Pooling (cpu, gpu cuda, gpu cuDNN)
URL: https://github.com/apache/incubator-mxnet/pull/13749#discussion_r253267148
##
File path: src/operator/nn/mkldnn/mkldnn_pooling-inl.h
##
@@ -104,7 +104,8 @@ class MKLDNNPoolingBwd {
inline bool SupportMKLDNNPooling(const PoolingParam ¶m) {
return param.kernel.ndim() == 2 &&
(param.pool_type == pool_enum::kMaxPooling ||
- param.pool_type == pool_enum::kAvgPooling);
+ param.pool_type == pool_enum::kAvgPooling) &&
+ (!param.layout.has_value() || param.layout.value() == mshadow::kNCHW);
Review comment:
Thank you for changing this. Just to clarify, MKL-DNN pooling primitive does
support NHWC input format. But when we integrated MKL-DNN backend into MXNet,
we assumed that user input data should always be NCHW for 4D tesnsor. We need
re-evaluated the workflow and integration to see if we need support NHWC here.
This is an automated message from the Apache Git Service.
To respond to the message, please log on 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] TaoLv commented on a change in pull request #13749: Add NHWC layout support to Pooling (cpu, gpu cuda, gpu cuDNN)
TaoLv commented on a change in pull request #13749: Add NHWC layout support to Pooling (cpu, gpu cuda, gpu cuDNN) URL: https://github.com/apache/incubator-mxnet/pull/13749#discussion_r253266876 ## File path: tests/python/gpu/test_operator_gpu.py ## @@ -608,6 +608,72 @@ def test_convolution_versions(): @with_seed() +def test_pooling_with_convention(): Review comment: Only max pooling is tested? This is an automated message from the Apache Git Service. To respond to the message, please log on 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] TaoLv commented on a change in pull request #13749: Add NHWC layout support to Pooling (cpu, gpu cuda, gpu cuDNN)
TaoLv commented on a change in pull request #13749: Add NHWC layout support to
Pooling (cpu, gpu cuda, gpu cuDNN)
URL: https://github.com/apache/incubator-mxnet/pull/13749#discussion_r253266821
##
File path: src/operator/nn/pooling.cc
##
@@ -421,11 +463,16 @@ NNVM_REGISTER_OP(_backward_Pooling)
.set_attr(
"FInplaceOption",
[](const NodeAttrs &attrs) {
-#if MXNET_USE_CUDNN == 1
- return std::vector >();
-#else
- return std::vector >{{1, 0}};
+#if MXNET_USE_MKLDNN == 1 && MXNET_USE_CUDA == 0 && MXNET_USE_CUDNN == 0
Review comment:
Ummm, do you have any case to validate this scenario? If MXNet is built with
both MKL-DNN and cuDNN enabled (eg. mxnet-cu90mkl package) and user gives
`ctx=cpu`. Is there any functionality or performance regression?
This is an automated message from the Apache Git Service.
To respond to the message, please log on 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. zhasheng 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 0755472 Bump the publish timestamp. 0755472 is described below commit 075547271591247ac99512b83e60cfc52be41da3 Author: mxnet-ci AuthorDate: Sat Feb 2 13:08:55 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..5bc02f7 --- /dev/null +++ b/date.txt @@ -0,0 +1 @@ +Sat Feb 2 13:08:55 UTC 2019
[GitHub] rongzha1 commented on a change in pull request #14056: A better split-2D(SliceChannel) op forward kernel for CPU
rongzha1 commented on a change in pull request #14056: A better
split-2D(SliceChannel) op forward kernel for CPU
URL: https://github.com/apache/incubator-mxnet/pull/14056#discussion_r253264609
##
File path: src/operator/channel_op_common.h
##
@@ -101,6 +101,42 @@ void Split(const mshadow::Tensor &input,
split_helper(input, output, dimension, req);
}
}
+
+template
+void Split_2D(const mshadow::Tensor &input,
+ std::vector > *output,
+ const int dimension, const std::vector &req) {
+ if (dimension != 1) {
+LOG(FATAL) << "dimension (" << dimension << ") must == 1";
+ }
+ if (dim != 3) {
Review comment:
same above
This is an automated message from the Apache Git Service.
To respond to the message, please log on 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] rongzha1 commented on a change in pull request #14056: A better split-2D(SliceChannel) op forward kernel for CPU
rongzha1 commented on a change in pull request #14056: A better
split-2D(SliceChannel) op forward kernel for CPU
URL: https://github.com/apache/incubator-mxnet/pull/14056#discussion_r253264597
##
File path: src/operator/channel_op_common.h
##
@@ -101,6 +101,42 @@ void Split(const mshadow::Tensor &input,
split_helper(input, output, dimension, req);
}
}
+
+template
+void Split_2D(const mshadow::Tensor &input,
+ std::vector > *output,
+ const int dimension, const std::vector &req) {
+ if (dimension != 1) {
Review comment:
Same coding style as functions in this file did such as Split(),
Concatenate()
This is an automated message from the Apache Git Service.
To respond to the message, please log on 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] ZhennanQin commented on a change in pull request #14056: A better split-2D(SliceChannel) op forward kernel for CPU
ZhennanQin commented on a change in pull request #14056: A better
split-2D(SliceChannel) op forward kernel for CPU
URL: https://github.com/apache/incubator-mxnet/pull/14056#discussion_r253263944
##
File path: src/operator/channel_op_common.h
##
@@ -101,6 +101,42 @@ void Split(const mshadow::Tensor &input,
split_helper(input, output, dimension, req);
}
}
+
+template
+void Split_2D(const mshadow::Tensor &input,
+ std::vector > *output,
+ const int dimension, const std::vector &req) {
+ if (dimension != 1) {
+LOG(FATAL) << "dimension (" << dimension << ") must == 1";
+ }
+ if (dim != 3) {
+LOG(FATAL) << "dimension (" << dim << ") must == 3";
+ } else {
+std::vector > out = *output;
+size_t size = out.size();
+std::vectorslice_len;
+std::vectorbegin_pos;
+begin_pos.push_back(0);
+
+for (index_t i = 0; i < size; ++i) {
+ slice_len.push_back(out[i].size(dimension));
+ begin_pos.push_back(begin_pos[i] + out[i].size(dimension));
+}
+#pragma omp parallel for
num_threads(engine::OpenMP::Get()->GetRecommendedOMPThreadCount())
+for (int i = 0; i < input.shape_[0]; i++) {
+ int iRow = i*input.shape_[1];
Review comment:
Add blank before and after operator.
This is an automated message from the Apache Git Service.
To respond to the message, please log on 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] ZhennanQin commented on a change in pull request #14056: A better split-2D(SliceChannel) op forward kernel for CPU
ZhennanQin commented on a change in pull request #14056: A better
split-2D(SliceChannel) op forward kernel for CPU
URL: https://github.com/apache/incubator-mxnet/pull/14056#discussion_r253262104
##
File path: src/operator/channel_op_common.h
##
@@ -101,6 +101,42 @@ void Split(const mshadow::Tensor &input,
split_helper(input, output, dimension, req);
}
}
+
+template
+void Split_2D(const mshadow::Tensor &input,
+ std::vector > *output,
+ const int dimension, const std::vector &req) {
+ if (dimension != 1) {
+LOG(FATAL) << "dimension (" << dimension << ") must == 1";
+ }
+ if (dim != 3) {
Review comment:
Same above.
This is an automated message from the Apache Git Service.
To respond to the message, please log on 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] ZhennanQin commented on a change in pull request #14056: A better split-2D(SliceChannel) op forward kernel for CPU
ZhennanQin commented on a change in pull request #14056: A better
split-2D(SliceChannel) op forward kernel for CPU
URL: https://github.com/apache/incubator-mxnet/pull/14056#discussion_r253262122
##
File path: src/operator/channel_op_common.h
##
@@ -101,6 +101,42 @@ void Split(const mshadow::Tensor &input,
split_helper(input, output, dimension, req);
}
}
+
+template
+void Split_2D(const mshadow::Tensor &input,
+ std::vector > *output,
+ const int dimension, const std::vector &req) {
+ if (dimension != 1) {
+LOG(FATAL) << "dimension (" << dimension << ") must == 1";
+ }
+ if (dim != 3) {
+LOG(FATAL) << "dimension (" << dim << ") must == 3";
+ } else {
+std::vector > out = *output;
+size_t size = out.size();
+std::vectorslice_len;
Review comment:
Add blank between type and variable.
This is an automated message from the Apache Git Service.
To respond to the message, please log on 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] ZhennanQin commented on a change in pull request #14056: A better split-2D(SliceChannel) op forward kernel for CPU
ZhennanQin commented on a change in pull request #14056: A better split-2D(SliceChannel) op forward kernel for CPU URL: https://github.com/apache/incubator-mxnet/pull/14056#discussion_r253262059 ## File path: src/operator/channel_op_common.h ## @@ -101,6 +101,42 @@ void Split(const mshadow::Tensor &input, split_helper(input, output, dimension, req); } } + +template +void Split_2D(const mshadow::Tensor &input, + std::vector > *output, Review comment: indent? This is an automated message from the Apache Git Service. To respond to the message, please log on 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] ZhennanQin commented on a change in pull request #14056: A better split-2D(SliceChannel) op forward kernel for CPU
ZhennanQin commented on a change in pull request #14056: A better
split-2D(SliceChannel) op forward kernel for CPU
URL: https://github.com/apache/incubator-mxnet/pull/14056#discussion_r253262094
##
File path: src/operator/channel_op_common.h
##
@@ -101,6 +101,42 @@ void Split(const mshadow::Tensor &input,
split_helper(input, output, dimension, req);
}
}
+
+template
+void Split_2D(const mshadow::Tensor &input,
+ std::vector > *output,
+ const int dimension, const std::vector &req) {
+ if (dimension != 1) {
Review comment:
Strange code style. Change to CHECK instead?
This is an automated message from the Apache Git Service.
To respond to the message, please log on 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] ZhennanQin commented on a change in pull request #14056: A better split-2D(SliceChannel) op forward kernel for CPU
ZhennanQin commented on a change in pull request #14056: A better
split-2D(SliceChannel) op forward kernel for CPU
URL: https://github.com/apache/incubator-mxnet/pull/14056#discussion_r253262164
##
File path: src/operator/channel_op_common.h
##
@@ -101,6 +101,42 @@ void Split(const mshadow::Tensor &input,
split_helper(input, output, dimension, req);
}
}
+
+template
+void Split_2D(const mshadow::Tensor &input,
+ std::vector > *output,
+ const int dimension, const std::vector &req) {
+ if (dimension != 1) {
+LOG(FATAL) << "dimension (" << dimension << ") must == 1";
+ }
+ if (dim != 3) {
+LOG(FATAL) << "dimension (" << dim << ") must == 3";
+ } else {
+std::vector > out = *output;
Review comment:
Seems redundant assignment. Why don't use output directly?
This is an automated message from the Apache Git Service.
To respond to the message, please log on 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] ZhennanQin commented on a change in pull request #14056: A better split-2D(SliceChannel) op forward kernel for CPU
ZhennanQin commented on a change in pull request #14056: A better
split-2D(SliceChannel) op forward kernel for CPU
URL: https://github.com/apache/incubator-mxnet/pull/14056#discussion_r253262215
##
File path: src/operator/channel_op_common.h
##
@@ -101,6 +101,42 @@ void Split(const mshadow::Tensor &input,
split_helper(input, output, dimension, req);
}
}
+
+template
+void Split_2D(const mshadow::Tensor &input,
+ std::vector > *output,
+ const int dimension, const std::vector &req) {
+ if (dimension != 1) {
+LOG(FATAL) << "dimension (" << dimension << ") must == 1";
+ }
+ if (dim != 3) {
+LOG(FATAL) << "dimension (" << dim << ") must == 3";
+ } else {
+std::vector > out = *output;
+size_t size = out.size();
+std::vectorslice_len;
+std::vectorbegin_pos;
+begin_pos.push_back(0);
+
+for (index_t i = 0; i < size; ++i) {
+ slice_len.push_back(out[i].size(dimension));
+ begin_pos.push_back(begin_pos[i] + out[i].size(dimension));
+}
+#pragma omp parallel for
num_threads(engine::OpenMP::Get()->GetRecommendedOMPThreadCount())
+for (int i = 0; i < input.shape_[0]; i++) {
+ int iRow = i*input.shape_[1];
+ for (int j = 0; j < size; j++) {
+int jRow = i*slice_len[j];
Review comment:
Add blank before and after operator. Please clean up all code style issue.
This is an automated message from the Apache Git Service.
To respond to the message, please log on 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] ZhennanQin commented on a change in pull request #14056: A better split-2D(SliceChannel) op forward kernel for CPU
ZhennanQin commented on a change in pull request #14056: A better
split-2D(SliceChannel) op forward kernel for CPU
URL: https://github.com/apache/incubator-mxnet/pull/14056#discussion_r253262126
##
File path: src/operator/channel_op_common.h
##
@@ -101,6 +101,42 @@ void Split(const mshadow::Tensor &input,
split_helper(input, output, dimension, req);
}
}
+
+template
+void Split_2D(const mshadow::Tensor &input,
+ std::vector > *output,
+ const int dimension, const std::vector &req) {
+ if (dimension != 1) {
+LOG(FATAL) << "dimension (" << dimension << ") must == 1";
+ }
+ if (dim != 3) {
+LOG(FATAL) << "dimension (" << dim << ") must == 3";
+ } else {
+std::vector > out = *output;
+size_t size = out.size();
+std::vectorslice_len;
+std::vectorbegin_pos;
Review comment:
Add blank between type and variable.
This is an automated message from the Apache Git Service.
To respond to the message, please log on 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] ZhennanQin commented on a change in pull request #14056: A better split-2D(SliceChannel) op forward kernel for CPU
ZhennanQin commented on a change in pull request #14056: A better
split-2D(SliceChannel) op forward kernel for CPU
URL: https://github.com/apache/incubator-mxnet/pull/14056#discussion_r253263942
##
File path: src/operator/channel_op_common.h
##
@@ -101,6 +101,42 @@ void Split(const mshadow::Tensor &input,
split_helper(input, output, dimension, req);
}
}
+
+template
+void Split_2D(const mshadow::Tensor &input,
+ std::vector > *output,
+ const int dimension, const std::vector &req) {
+ if (dimension != 1) {
+LOG(FATAL) << "dimension (" << dimension << ") must == 1";
+ }
+ if (dim != 3) {
+LOG(FATAL) << "dimension (" << dim << ") must == 3";
+ } else {
+std::vector > out = *output;
+size_t size = out.size();
+std::vectorslice_len;
+std::vectorbegin_pos;
+begin_pos.push_back(0);
+
+for (index_t i = 0; i < size; ++i) {
+ slice_len.push_back(out[i].size(dimension));
+ begin_pos.push_back(begin_pos[i] + out[i].size(dimension));
+}
+#pragma omp parallel for
num_threads(engine::OpenMP::Get()->GetRecommendedOMPThreadCount())
Review comment:
Is it possible to parallel both 2 outer loops? Try collapse keyword. And
beware that it's only works on Linux platform. You can find the example code in
https://github.com/apache/incubator-mxnet/blob/master/src/operator/l2_normalization.cc
This is an automated message from the Apache Git Service.
To respond to the message, please log on 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] bputrycz commented on issue #10715: Wrong params file is not reported as error in C API
bputrycz commented on issue #10715: Wrong params file is not reported as error in C API URL: https://github.com/apache/incubator-mxnet/issues/10715#issuecomment-459955392 I wasn't blocked by this issue, just noticed it and reported. I haven't invested more time into it. I haven't tested it with the newest MXNet, and I don't want to do it now. It is your call now to decide if it is worth to check again. I think, I put above easily replicable code: python for generation of the sample model and C code with prediction from it. It misses only main() and includes. Then to see the problem, you need to modify params file with changing the names of the keys in the dictionary. This is an automated message from the Apache Git Service. To respond to the message, please log on 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] arcadiaphy opened a new pull request #14058: add backgroud class in box_nms
arcadiaphy opened a new pull request #14058: add backgroud class in box_nms URL: https://github.com/apache/incubator-mxnet/pull/14058 ## Description ## This PR is mentioned in #14057 What I have done in box_nms operator: 1. add background_id argument 2. filter out background boxes before sorting operation ## 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 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] arcadiaphy opened a new issue #14057: validation stucks when training gluoncv ssd model
arcadiaphy opened a new issue #14057: validation stucks when training gluoncv ssd model URL: https://github.com/apache/incubator-mxnet/issues/14057 ## Description When training gluoncv ssd model, validation sometimes takes way more longer time than the training epoch. After debugging, the problem comes from the `box_nms` operator which contributes most of the time. ## Environment info (Required) ``` Centos 7 CUDA: 9.0 cudnn: 7 mxnet: 1.4.0.rc2 gluon-cv: latest ``` ## Minimum reproducible example The following snippets show `box_nms` will take very long time when processing a lot of prior boxes ``` import mxnet as mx import numpy as np np.random.seed(0) batch_size = 32 prior_number = 10 data = np.zeros((batch_size, prior_number, 6)) data[:, :, 0] = np.random.randint(-1, 1, (batch_size, prior_number)) data[:, :, 1] = np.random.random((batch_size, prior_number)) xmin = np.random.random((batch_size, prior_number)) ymin = np.random.random((batch_size, prior_number)) width = np.random.random((batch_size, prior_number)) height = np.random.random((batch_size, prior_number)) data[:, :, 2] = xmin data[:, :, 3] = ymin data[:, :, 4] = xmin + width data[:, :, 5] = ymin + height mx_data = mx.nd.array(data, ctx=mx.gpu(0)) rv = mx.nd.contrib.box_nms(mx_data, overlap_thresh=0.5, valid_thresh=0.01, topk=400, score_index=1, id_index=0) mx.nd.waitall() ``` ## What I have found out 1. The gpu version of stable sort in `SortByKey` function degrades badly on sorting length 2. The `box_nms` operator doesn't remove background boxes in valid box filtering which leads to big sorting length ## What I have done 1. Add SORT_WITH_THRUST compiling definition in Makefile: the validation process is still very slow 2. Add background boxes filtering in `box_nms`: the validation process accelerates dramatically since most of boxes are classified as background. I will post a PR on the second solution. This is an automated message from the Apache Git Service. To respond to the message, please log on 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] ZhennanQin commented on issue #14053: in-place reshape ops
ZhennanQin commented on issue #14053: in-place reshape ops URL: https://github.com/apache/incubator-mxnet/pull/14053#issuecomment-459949587 Have you tried this with mkldnn enabled? E.g reshape the output of mkldnn convolution? This is an automated message from the Apache Git Service. To respond to the message, please log on 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
