split tensor operator files (apache#4150)

* fix examples * fix * split tensor operator files
piiswrong · Dec 29, 2016 · adb7cb1 · adb7cb1
1 parent 2be08f0
commit adb7cb1
Show file tree

Hide file tree

Showing 25 changed files with 312 additions and 215 deletions.
diff --git a/example/nce-loss/lstm_word.py b/example/nce-loss/lstm_word.py
@@ -1,4 +1,5 @@
 # pylint:skip-file
+import logging
 import sys, random, time, math
 sys.path.insert(0, "../../python")
 import mxnet as mx
@@ -182,6 +183,9 @@ def reset(self):
         pass
 
 if __name__ == '__main__':
+    head = '%(asctime)-15s %(message)s'
+    logging.basicConfig(level=logging.DEBUG, format=head)
+
     parser = OptionParser()
     parser.add_option("-g", "--gpu", action = "store_true", dest = "gpu", default = False,
                       help = "use gpu")
@@ -195,6 +199,7 @@ def reset(self):
     init_h = [('l%d_init_h'%l, (batch_size, num_hidden)) for l in range(num_lstm_layer)]
     init_states = init_c + init_h
 
+
     data_train = DataIter("./data/text8", batch_size, seq_len, num_label,
                           init_states)
 
@@ -210,10 +215,6 @@ def reset(self):
                                  momentum = 0.9,
                                  wd = 0.0000,
                                  initializer=mx.init.Xavier(factor_type="in", magnitude=2.34))
-
-    import logging
-    head = '%(asctime)-15s %(message)s'
-    logging.basicConfig(level=logging.DEBUG, format=head)
 
     metric = NceLSTMAuc()
     model.fit(X = data_train,

diff --git a/example/nce-loss/toy_nce.py b/example/nce-loss/toy_nce.py
@@ -1,4 +1,5 @@
 # pylint:skip-file
+import logging
 import sys, random, time
 sys.path.insert(0, "../../python")
 import mxnet as mx
@@ -83,6 +84,9 @@ def reset(self):
         pass
 
 if __name__ == '__main__':
+    head = '%(asctime)-15s %(message)s'
+    logging.basicConfig(level=logging.DEBUG, format=head)
+
     batch_size = 128
     vocab_size = 10000
     feature_size = 100
@@ -100,9 +104,6 @@ def reset(self):
                                  momentum = 0.9,
                                  wd = 0.00001,
                                  initializer=mx.init.Xavier(factor_type="in", magnitude=2.34))
-    import logging
-    head = '%(asctime)-15s %(message)s'
-    logging.basicConfig(level=logging.DEBUG, format=head)
 
     metric = NceAccuracy()
     model.fit(X = data_train, eval_data = data_test,

diff --git a/example/nce-loss/toy_softmax.py b/example/nce-loss/toy_softmax.py
@@ -1,4 +1,5 @@
 # pylint:skip-file
+import logging
 import sys, random, time
 sys.path.insert(0, "../../python")
 import mxnet as mx
@@ -72,11 +73,14 @@ def reset(self):
         pass
 
 if __name__ == '__main__':
+    head = '%(asctime)-15s %(message)s'
+    logging.basicConfig(level=logging.DEBUG, format=head)
+
     batch_size = 128
     vocab_size = 10000
     feature_size = 100
     num_label = 6
-    
+
     data_train = DataIter(100000, batch_size, vocab_size, num_label, feature_size)
     data_test = DataIter(1000, batch_size, vocab_size, num_label, feature_size)
 
@@ -89,9 +93,6 @@ def reset(self):
                                  momentum = 0.9,
                                  wd = 0.0000,
                                  initializer=mx.init.Xavier(factor_type="in", magnitude=2.34))
-    import logging
-    head = '%(asctime)-15s %(message)s'
-    logging.basicConfig(level=logging.DEBUG, format=head)
 
     model.fit(X = data_train, eval_data = data_test,
               batch_end_callback = mx.callback.Speedometer(batch_size, 50),)

diff --git a/example/nce-loss/wordvec.py b/example/nce-loss/wordvec.py
@@ -1,4 +1,5 @@
 # pylint:skip-file
+import logging
 import sys, random, time, math
 sys.path.insert(0, "../../python")
 import mxnet as mx
@@ -116,6 +117,9 @@ def reset(self):
         pass
 
 if __name__ == '__main__':
+    head = '%(asctime)-15s %(message)s'
+    logging.basicConfig(level=logging.DEBUG, format=head)
+
     parser = OptionParser()
     parser.add_option("-g", "--gpu", action = "store_true", dest = "gpu", default = False,
                       help = "use gpu")
@@ -138,9 +142,6 @@ def reset(self):
                                  wd = 0.0000,
                                  initializer=mx.init.Xavier(factor_type="in", magnitude=2.34))
 
-    import logging
-    head = '%(asctime)-15s %(message)s'
-    logging.basicConfig(level=logging.DEBUG, format=head)
 
     metric = NceAuc()
     model.fit(X = data_train,

diff --git a/example/nce-loss/wordvec_subwords.py b/example/nce-loss/wordvec_subwords.py
@@ -1,4 +1,5 @@
 # pylint:skip-file
+import logging
 import sys, random, time, math
 import mxnet as mx
 import numpy as np
@@ -247,6 +248,9 @@ def reset(self):
 
 
 if __name__ == '__main__':
+    head = '%(asctime)-15s %(message)s'
+    logging.basicConfig(level=logging.DEBUG, format=head)
+
     parser = OptionParser()
     parser.add_option("-g", "--gpu", action="store_true", dest="gpu", default=False,
                       help="use gpu")
@@ -271,11 +275,6 @@ def reset(self):
                                  wd=0.0000,
                                  initializer=mx.init.Xavier(factor_type="in", magnitude=2.34))
 
-    import logging
-
-    head = '%(asctime)-15s %(message)s'
-    logging.basicConfig(level=logging.DEBUG, format=head)
-
     metric = NceAuc()
     model.fit(X=data_train,
               eval_metric=metric,

diff --git a/example/rnn-time-major/get_ptb_data.sh b/example/rnn-time-major/get_ptb_data.sh
@@ -0,0 +1,14 @@
+#!/usr/bin/env bash
+
+RNN_DIR=$(cd `dirname $0`; pwd)
+DATA_DIR="${RNN_DIR}/data/"
+
+if [[ ! -d "${DATA_DIR}" ]]; then
+  echo "${DATA_DIR} doesn't exist, will create one";
+  mkdir -p ${DATA_DIR}
+fi
+
+wget -P ${DATA_DIR} https://raw.githubusercontent.com/dmlc/web-data/master/mxnet/ptb/ptb.train.txt;
+wget -P ${DATA_DIR} https://raw.githubusercontent.com/dmlc/web-data/master/mxnet/ptb/ptb.valid.txt;
+wget -P ${DATA_DIR} https://raw.githubusercontent.com/dmlc/web-data/master/mxnet/ptb/ptb.test.txt;
+wget -P ${DATA_DIR} https://raw.githubusercontent.com/dmlc/web-data/master/mxnet/tinyshakespeare/input.txt;
diff --git a/example/rnn/README.md b/example/rnn/README.md
@@ -2,6 +2,9 @@ RNN Example
 ===========
 This folder contains RNN examples using low level symbol interface.
 
+## Data
+Run `get_ptb_data.sh` to download PenTreeBank data.
+
 ## Python
 
 - [lstm.py](lstm.py) Functions for building a LSTM Network

diff --git a/nnvm b/nnvm
diff --git a/python/mxnet/model.py b/python/mxnet/model.py
@@ -428,8 +428,9 @@ def __init__(self, symbol, ctx=None,
                  allow_extra_params=False,
                  begin_epoch=0,
                  **kwargs):
-        logging.warning('[Deprecation Warning] mxnet.model.FeedForward has been deprecated. ' + \
-                        'Please use mxnet.mod.Module instead.')
+        logging.warning(
+            '\033[91m[Deprecation Warning] mxnet.model.FeedForward has been deprecated. ' + \
+            'Please use mxnet.mod.Module instead.\033[0m')
 
         if isinstance(symbol, sym.Symbol):
             self.symbol = symbol

diff --git a/python/mxnet/symbol.py b/python/mxnet/symbol.py
@@ -44,8 +44,9 @@ class Symbol(SymbolBase):
 
     def __repr__(self):
         """Get a string representation of the symbol."""
+        name = self.name
         return '<%s %s>' % (self.__class__.__name__,
-                            self.name)
+                            'Grouped' if name is None else name)
 
     def __add__(self, other):
         if isinstance(other, Symbol):

diff --git a/src/operator/elemwise_op_common.h b/src/operator/elemwise_op_common.h
@@ -74,7 +74,7 @@ template<int n_in, int n_out>
 inline bool ElemwiseShape(const nnvm::NodeAttrs& attrs,
                           std::vector<TShape> *in_attrs,
                           std::vector<TShape> *out_attrs) {
-  CHECK_EQ(in_attrs->size(), n_in) << attrs.name << in_attrs->size() << n_in;
+  CHECK_EQ(in_attrs->size(), n_in) << " in operator " << attrs.name;
   CHECK_EQ(out_attrs->size(), n_out);
   return ElemwiseAttr<TShape, shape_is_none, true>(
     attrs, in_attrs, out_attrs);
@@ -88,7 +88,7 @@ template<int n_in, int n_out>
 inline bool ElemwiseType(const nnvm::NodeAttrs& attrs,
                          std::vector<int> *in_attrs,
                          std::vector<int> *out_attrs) {
-  CHECK_EQ(in_attrs->size(), n_in) << attrs.name << in_attrs->size() << n_in;
+  CHECK_EQ(in_attrs->size(), n_in) << " in operator " << attrs.name;
   CHECK_EQ(out_attrs->size(), n_out);
   return ElemwiseAttr<int, type_is_none, true>(
     attrs, in_attrs, out_attrs);

diff --git a/src/operator/loss_binary_op.cc b/src/operator/loss_binary_op.cc
@@ -13,7 +13,7 @@ NNVM_REGISTER_OP(softmax_cross_entropy)
 .set_num_inputs(2)
 .set_num_outputs(1)
 .set_attr<nnvm::FInferShape>("FInferShape", SoftmaxCrossEntropyShape)
-.set_attr<nnvm::FInferType>("FInferType", ElemwiseType<1, 1>)
+.set_attr<nnvm::FInferType>("FInferType", ElemwiseType<2, 1>)
 .set_attr<FResourceRequest>("FResourceRequest",
   [](const NodeAttrs& attrs) {
     return std::vector<ResourceRequest>{ResourceRequest::kTempSpace};

diff --git a/src/operator/tensor/broadcast_reduce_op_index.cc b/src/operator/tensor/broadcast_reduce_op_index.cc
@@ -0,0 +1,34 @@
+/*!
+ *  Copyright (c) 2016 by Contributors
+ * \file broadcast_reduce_op.cc
+ * \brief CPU Implementation of broadcast and reduce functions.
+ */
+#include "./broadcast_reduce_op.h"
+
+namespace mxnet {
+namespace op {
+MXNET_OPERATOR_REGISTER_REDUCE_AXIS(argmax)
+.MXNET_DESCRIBE("Compute argmax")
+.set_attr<FCompute>("FCompute<cpu>", SearchAxisCompute<cpu, mshadow::red::maximum>);
+
+MXNET_OPERATOR_REGISTER_REDUCE_AXIS(argmin)
+.MXNET_DESCRIBE("Compute argmin")
+.set_attr<FCompute>("FCompute<cpu>", SearchAxisCompute<cpu, mshadow::red::minimum>);
+
+// Legacy support
+NNVM_REGISTER_OP(argmax_channel)
+.set_num_inputs(1)
+.set_num_outputs(1)
+.set_attr_parser([](NodeAttrs* attrs) {
+    ReduceAxisParam param;
+    param.axis = 1;
+    param.keepdims = false;
+    attrs->parsed = param;
+  })
+.set_attr<nnvm::FInferShape>("FInferShape", ReduceAxisShape)
+.set_attr<nnvm::FInferType>("FInferType", ElemwiseType<1, 1>)
+.set_attr<FCompute>("FCompute<cpu>", SearchAxisCompute<cpu, mshadow::red::maximum>)
+.add_argument("src", "NDArray", "Source input");
+
+}  // namespace op
+}  // namespace mxnet
diff --git a/src/operator/tensor/broadcast_reduce_op_index.cu b/src/operator/tensor/broadcast_reduce_op_index.cu
@@ -0,0 +1,21 @@
+/*!
+ *  Copyright (c) 2016 by Contributors
+ * \file broadcast_reduce_op.cu
+ * \brief GPU Implementation of broadcast and reduce functions.
+ */
+#include "./broadcast_reduce_op.h"
+
+namespace mxnet {
+namespace op {
+NNVM_REGISTER_OP(argmax)
+.set_attr<FCompute>("FCompute<gpu>", SearchAxisCompute<gpu, mshadow::red::maximum>);
+
+NNVM_REGISTER_OP(argmin)
+.set_attr<FCompute>("FCompute<gpu>", SearchAxisCompute<gpu, mshadow::red::minimum>);
+
+// Legacy support
+NNVM_REGISTER_OP(argmax_channel)
+.set_attr<FCompute>("FCompute<gpu>", SearchAxisCompute<gpu, mshadow::red::maximum>);
+
+}  // namespace op
+}  // namespace mxnet
diff --git a/src/operator/tensor/broadcast_reduce_op.cc → ...rator/tensor/broadcast_reduce_op_value.cc b/src/operator/tensor/broadcast_reduce_op.cc → ...rator/tensor/broadcast_reduce_op_value.cc
@@ -62,29 +62,6 @@ NNVM_REGISTER_OP(_broadcast_backward)
 .set_attr<nnvm::TIsBackward>("TIsBackward", true)
 .set_attr<FCompute>("FCompute<cpu>", ReduceAxesCompute<cpu, mshadow::red::sum>);
 
-MXNET_OPERATOR_REGISTER_REDUCE_AXIS(argmax)
-.MXNET_DESCRIBE("Compute argmax")
-.set_attr<FCompute>("FCompute<cpu>", SearchAxisCompute<cpu, mshadow::red::maximum>);
-
-MXNET_OPERATOR_REGISTER_REDUCE_AXIS(argmin)
-.MXNET_DESCRIBE("Compute argmin")
-.set_attr<FCompute>("FCompute<cpu>", SearchAxisCompute<cpu, mshadow::red::minimum>);
-
-// Legacy support
-NNVM_REGISTER_OP(argmax_channel)
-.set_num_inputs(1)
-.set_num_outputs(1)
-.set_attr_parser([](NodeAttrs* attrs) {
-    ReduceAxisParam param;
-    param.axis = 1;
-    param.keepdims = false;
-    attrs->parsed = param;
-  })
-.set_attr<nnvm::FInferShape>("FInferShape", ReduceAxisShape)
-.set_attr<nnvm::FInferType>("FInferType", ElemwiseType<1, 1>)
-.set_attr<FCompute>("FCompute<cpu>", SearchAxisCompute<cpu, mshadow::red::maximum>)
-.add_argument("src", "NDArray", "Source input");
-
 NNVM_REGISTER_OP(norm)
 .set_num_inputs(1)
 .set_num_outputs(1)

diff --git a/src/operator/tensor/broadcast_reduce_op.cu → ...rator/tensor/broadcast_reduce_op_value.cu b/src/operator/tensor/broadcast_reduce_op.cu → ...rator/tensor/broadcast_reduce_op_value.cu
@@ -34,16 +34,6 @@ NNVM_REGISTER_OP(broadcast_to)
 NNVM_REGISTER_OP(_broadcast_backward)
 .set_attr<FCompute>("FCompute<gpu>", ReduceAxesCompute<gpu, mshadow::red::sum>);
 
-NNVM_REGISTER_OP(argmax)
-.set_attr<FCompute>("FCompute<gpu>", SearchAxisCompute<gpu, mshadow::red::maximum>);
-
-NNVM_REGISTER_OP(argmin)
-.set_attr<FCompute>("FCompute<gpu>", SearchAxisCompute<gpu, mshadow::red::minimum>);
-
-// Legacy support
-NNVM_REGISTER_OP(argmax_channel)
-.set_attr<FCompute>("FCompute<gpu>", SearchAxisCompute<gpu, mshadow::red::maximum>);
-
 NNVM_REGISTER_OP(norm)
 .set_attr<FCompute>("FCompute<gpu>", L2NormCompute<gpu>);