From 01c16d6c76baded5f8eaea5d66d7f54bcb20af1d Mon Sep 17 00:00:00 2001
From: Christian Fuss <chrifuss@freenet.de>
Date: Mon, 9 Sep 2019 15:59:46 +0200
Subject: [PATCH] added functionality for ArgMax layer
---
.../templates/gluon/CNNSupervisedTrainer.ftl | 35 ++++++++++++-------
.../templates/gluon/elements/ArgMax.ftl | 3 +-
.../templates/gluon/pythonExecute.ftl | 10 ++++++
.../CNNSupervisedTrainer_Alexnet.py | 26 +++++++++-----
...upervisedTrainer_CifarClassifierNetwork.py | 26 +++++++++-----
.../target_code/CNNSupervisedTrainer_VGG16.py | 26 +++++++++-----
6 files changed, 89 insertions(+), 37 deletions(-)
diff --git a/src/main/resources/templates/gluon/CNNSupervisedTrainer.ftl b/src/main/resources/templates/gluon/CNNSupervisedTrainer.ftl
index c2bb59b9..579ac61b 100644
--- a/src/main/resources/templates/gluon/CNNSupervisedTrainer.ftl
+++ b/src/main/resources/templates/gluon/CNNSupervisedTrainer.ftl
@@ -185,13 +185,19 @@ class ${tc.fileNameWithoutEnding}:
if True: <#-- Fix indentation -->
<#include "pythonExecute.ftl">
- predictions = [
-<#list tc.architectureOutputs as output_name>
- mx.nd.argmax(${output_name}, axis=1)<#sep>,
-</#list>
-]
+ out_names=[]
+ <#list tc.architectureOutputs as output_name>
+ out_names.append(${output_name})
+ </#list>
+ predictions = []
+ for output_name in out_names:
+ if mx.nd.shape_array(output_name).size > 1:
+ predictions.append(mx.nd.argmax(output_name, axis=1))
+ #ArgMax already applied
+ else:
+ predictions.append(output_name)
- <#include "elements/BeamSearchStart.ftl">
+ <#include "elements/BeamSearch.ftl">
metric.update(preds=predictions, labels=labels)
train_metric_score = metric.get()[1]
@@ -213,12 +219,17 @@ class ${tc.fileNameWithoutEnding}:
if True: <#-- Fix indentation -->
<#include "pythonExecute.ftl">
- predictions = [
-<#list tc.architectureOutputs as output_name>
- mx.nd.argmax(${output_name}, axis=1)<#sep>,
-</#list>
-
- ]
+ out_names=[]
+ <#list tc.architectureOutputs as output_name>
+ out_names.append(${output_name})
+ </#list>
+ predictions = []
+ for output_name in out_names:
+ if mx.nd.shape_array(output_name).size > 1:
+ predictions.append(mx.nd.argmax(output_name, axis=1))
+ #ArgMax already applied
+ else:
+ predictions.append(output_name)
metric.update(preds=predictions, labels=labels)
test_metric_score = metric.get()[1]
diff --git a/src/main/resources/templates/gluon/elements/ArgMax.ftl b/src/main/resources/templates/gluon/elements/ArgMax.ftl
index 44c6e7ab..b6868735 100644
--- a/src/main/resources/templates/gluon/elements/ArgMax.ftl
+++ b/src/main/resources/templates/gluon/elements/ArgMax.ftl
@@ -1,4 +1,5 @@
<#assign input = element.inputs[0]>
<#if mode == "FORWARD_FUNCTION">
- ${element.name} = F.ndarray.argmax(${input}, keepdims=True)
+ <#-- only passtrough method, argmax logic is applied in pythonExecute.ftl and CNNSupervisedTrainer.ftl -->
+ ${element.name} = ${input}
</#if>
\ No newline at end of file
diff --git a/src/main/resources/templates/gluon/pythonExecute.ftl b/src/main/resources/templates/gluon/pythonExecute.ftl
index 40f86d65..ba627a11 100644
--- a/src/main/resources/templates/gluon/pythonExecute.ftl
+++ b/src/main/resources/templates/gluon/pythonExecute.ftl
@@ -9,10 +9,20 @@
<#if networkInstruction.isUnroll()>
<#list networkInstruction.toUnrollInstruction().resolvedBodies as resolvedBody>
${tc.join(tc.getStreamOutputNames(networkInstruction.body, resolvedBody), ", ")} = self._networks[${networkInstruction?index}](${tc.join(tc.getStreamInputNames(networkInstruction.body, resolvedBody), ", ")})
+ <#list resolvedBody.elements as element>
+ <#if element.name == "ArgMax">
+ ${tc.join(tc.getStreamOutputNames(networkInstruction.body, resolvedBody), ", ")} = mx.nd.argmax(${tc.join(tc.getStreamOutputNames(networkInstruction.body, resolvedBody), ", ")}, axis=1)
+ </#if>
+ </#list>
</#list>
<#else>
<#if networkInstruction.body.isTrainable()>
${tc.join(tc.getStreamOutputNames(networkInstruction.body), ", ")} = self._networks[${networkInstruction?index}](${tc.join(tc.getStreamInputNames(networkInstruction.body), ", ")})
+ <#list networkInstruction.body.elements as element>
+ <#if element.name == "ArgMax">
+ ${tc.join(tc.getStreamOutputNames(networkInstruction.body), ", ")} = mx.nd.argmax(${tc.join(tc.getStreamOutputNames(networkInstruction.body), ", ")}, axis=1)
+ </#if>
+ </#list>
<#else>
${tc.include(networkInstruction.body, "PYTHON_INLINE")}
</#if>
diff --git a/src/test/resources/target_code/CNNSupervisedTrainer_Alexnet.py b/src/test/resources/target_code/CNNSupervisedTrainer_Alexnet.py
index 795f867d..e3682957 100644
--- a/src/test/resources/target_code/CNNSupervisedTrainer_Alexnet.py
+++ b/src/test/resources/target_code/CNNSupervisedTrainer_Alexnet.py
@@ -140,7 +140,6 @@ class CNNSupervisedTrainer_Alexnet:
predictions_ = self._networks[0](data_)
-
loss = \
loss_function(predictions_, predictions_label)
@@ -178,9 +177,15 @@ class CNNSupervisedTrainer_Alexnet:
predictions_ = self._networks[0](data_)
-
- predictions = [
- mx.nd.argmax(predictions_, axis=1)]
+ out_names=[]
+ out_names.append(predictions_)
+ predictions = []
+ for output_name in out_names:
+ if mx.nd.shape_array(output_name).size > 1:
+ predictions.append(mx.nd.argmax(output_name, axis=1))
+ #ArgMax already applied
+ else:
+ predictions.append(output_name)
metric.update(preds=predictions, labels=labels)
@@ -200,10 +205,15 @@ class CNNSupervisedTrainer_Alexnet:
predictions_ = self._networks[0](data_)
-
- predictions = [
- mx.nd.argmax(predictions_, axis=1)
- ]
+ out_names=[]
+ out_names.append(predictions_)
+ predictions = []
+ for output_name in out_names:
+ if mx.nd.shape_array(output_name).size > 1:
+ predictions.append(mx.nd.argmax(output_name, axis=1))
+ #ArgMax already applied
+ else:
+ predictions.append(output_name)
metric.update(preds=predictions, labels=labels)
test_metric_score = metric.get()[1]
diff --git a/src/test/resources/target_code/CNNSupervisedTrainer_CifarClassifierNetwork.py b/src/test/resources/target_code/CNNSupervisedTrainer_CifarClassifierNetwork.py
index a34bcc25..7cf9b09a 100644
--- a/src/test/resources/target_code/CNNSupervisedTrainer_CifarClassifierNetwork.py
+++ b/src/test/resources/target_code/CNNSupervisedTrainer_CifarClassifierNetwork.py
@@ -140,7 +140,6 @@ class CNNSupervisedTrainer_CifarClassifierNetwork:
softmax_ = self._networks[0](data_)
-
loss = \
loss_function(softmax_, softmax_label)
@@ -178,9 +177,15 @@ class CNNSupervisedTrainer_CifarClassifierNetwork:
softmax_ = self._networks[0](data_)
-
- predictions = [
- mx.nd.argmax(softmax_, axis=1)]
+ out_names=[]
+ out_names.append(softmax_)
+ predictions = []
+ for output_name in out_names:
+ if mx.nd.shape_array(output_name).size > 1:
+ predictions.append(mx.nd.argmax(output_name, axis=1))
+ #ArgMax already applied
+ else:
+ predictions.append(output_name)
metric.update(preds=predictions, labels=labels)
@@ -200,10 +205,15 @@ class CNNSupervisedTrainer_CifarClassifierNetwork:
softmax_ = self._networks[0](data_)
-
- predictions = [
- mx.nd.argmax(softmax_, axis=1)
- ]
+ out_names=[]
+ out_names.append(softmax_)
+ predictions = []
+ for output_name in out_names:
+ if mx.nd.shape_array(output_name).size > 1:
+ predictions.append(mx.nd.argmax(output_name, axis=1))
+ #ArgMax already applied
+ else:
+ predictions.append(output_name)
metric.update(preds=predictions, labels=labels)
test_metric_score = metric.get()[1]
diff --git a/src/test/resources/target_code/CNNSupervisedTrainer_VGG16.py b/src/test/resources/target_code/CNNSupervisedTrainer_VGG16.py
index 290db22b..49090160 100644
--- a/src/test/resources/target_code/CNNSupervisedTrainer_VGG16.py
+++ b/src/test/resources/target_code/CNNSupervisedTrainer_VGG16.py
@@ -140,7 +140,6 @@ class CNNSupervisedTrainer_VGG16:
predictions_ = self._networks[0](data_)
-
loss = \
loss_function(predictions_, predictions_label)
@@ -178,9 +177,15 @@ class CNNSupervisedTrainer_VGG16:
predictions_ = self._networks[0](data_)
-
- predictions = [
- mx.nd.argmax(predictions_, axis=1)]
+ out_names=[]
+ out_names.append(predictions_)
+ predictions = []
+ for output_name in out_names:
+ if mx.nd.shape_array(output_name).size > 1:
+ predictions.append(mx.nd.argmax(output_name, axis=1))
+ #ArgMax already applied
+ else:
+ predictions.append(output_name)
metric.update(preds=predictions, labels=labels)
@@ -200,10 +205,15 @@ class CNNSupervisedTrainer_VGG16:
predictions_ = self._networks[0](data_)
-
- predictions = [
- mx.nd.argmax(predictions_, axis=1)
- ]
+ out_names=[]
+ out_names.append(predictions_)
+ predictions = []
+ for output_name in out_names:
+ if mx.nd.shape_array(output_name).size > 1:
+ predictions.append(mx.nd.argmax(output_name, axis=1))
+ #ArgMax already applied
+ else:
+ predictions.append(output_name)
metric.update(preds=predictions, labels=labels)
test_metric_score = metric.get()[1]
--
GitLab