Fix Issue #903 (#916)

* update readme in ga_squad

* update readme

* fix typo

* Update README.md

* Update README.md

* Update README.md

* update readme

* update

* fix path

* update reference

* fix bug in config file

* update nni_arch_overview.png

* update

* update

* update

* update home page

* fix issue: #902, random parameter

* update doc about random

* change Random tuner test

* update metric_test for Random tuner

* update random config in test

* update Evolution_SQuAD doc

docs/en_US/Builtin_Tuner.md

@@ -73,8 +73,6 @@ Random search is suggested when each trial does not take too long (e.g., each tr
 # config.yml
 tuner:
   builtinTunerName: Random
-  classArgs:
-    optimize_mode: maximize
 ```
 
 <br>
@@ -115,10 +113,6 @@ tuner:
 
 Its requirement of computation resource is relatively high. Specifically, it requires large initial population to avoid falling into local optimum. If your trial is short or leverages assessor, this tuner is a good choice. And, it is more suggested when your trial code supports weight transfer, that is, the trial could inherit the converged weights from its parent(s). This can greatly speed up the training progress.
 
-**Requirement of classArg**
-
-* **optimize_mode** (*maximize or minimize, optional, default = maximize*) - If 'maximize', tuners will return the hyperparameter set with larger expectation. If 'minimize', tuner will return the hyperparameter set with smaller expectation.
-
 **Usage example**
 
 ```yaml

docs/en_US/SQuAD_evolution_examples.md

@@ -2,8 +2,7 @@
 This example shows us how to use Genetic Algorithm to find good model architectures for Reading Comprehension.
 
 ## 1. Search Space
-Since attention and recurrent neural network (RNN) have been proven effective in Reading Comprehension.
-We conclude the search space as follow:
+Since attention and RNN have been proven effective in Reading Comprehension, we conclude the search space as follow:
 
 1. IDENTITY (Effectively means keep training).
 2. INSERT-RNN-LAYER (Inserts a LSTM. Comparing the performance of GRU and LSTM in our experiment, we decided to use LSTM here.)

src/sdk/pynni/nni/hyperopt_tuner/hyperopt_tuner.py

@@ -158,7 +158,7 @@ class HyperoptTuner(Tuner):
     HyperoptTuner is a tuner which using hyperopt algorithm.
     """
 
-    def __init__(self, algorithm_name, optimize_mode):
+    def __init__(self, algorithm_name, optimize_mode = 'minimize'):
         """
         Parameters
         ----------

test/config_test/examples/cifar10-pytorch.test.yml

@@ -7,8 +7,6 @@ searchSpacePath: ./cifar10_search_space.json
 
 tuner:
   builtinTunerName: Random
-  classArgs:
-    optimize_mode: maximize
 assessor:
   builtinAssessorName: Medianstop
   classArgs:

test/config_test/examples/mnist-annotation.test.yml

@@ -6,8 +6,6 @@ trialConcurrency: 2
 
 tuner:
   builtinTunerName: Random
-  classArgs:
-    optimize_mode: maximize
 assessor:
   builtinAssessorName: Medianstop
   classArgs:

test/config_test/examples/mnist-keras.test.yml

@@ -7,8 +7,6 @@ searchSpacePath: ../../../examples/trials/mnist-keras/search_space.json
 
 tuner:
   builtinTunerName: Random
-  classArgs:
-    optimize_mode: maximize
 assessor:
   builtinAssessorName: Medianstop
   classArgs:

test/config_test/examples/mnist.test.yml

@@ -7,8 +7,6 @@ searchSpacePath: ./mnist_search_space.json
 
 tuner:
   builtinTunerName: Random
-  classArgs:
-    optimize_mode: maximize
 assessor:
   builtinAssessorName: Medianstop
   classArgs:

test/config_test/examples/sklearn-classification.test.yml

@@ -7,8 +7,6 @@ searchSpacePath: ../../../examples/trials/sklearn/classification/search_space.js
 
 tuner:
   builtinTunerName: Random
-  classArgs:
-    optimize_mode: maximize
 assessor:
   builtinAssessorName: Medianstop
   classArgs:

test/config_test/examples/sklearn-regression.test.yml

@@ -7,8 +7,6 @@ searchSpacePath: ../../../examples/trials/sklearn/regression/search_space.json
 
 tuner:
   builtinTunerName: Random
-  classArgs:
-    optimize_mode: maximize
 assessor:
   builtinAssessorName: Medianstop
   classArgs:

test/config_test/tuners/mnist-annotation-random.test.yml

@@ -6,8 +6,6 @@ trialConcurrency: 1
 
 tuner:
   builtinTunerName: Random
-  classArgs:
-    optimize_mode: maximize
 assessor:
   builtinAssessorName: Medianstop
   classArgs:

test/metrics_test/metrics.test.yml

@@ -7,8 +7,6 @@ searchSpacePath: ./search_space.json
 
 tuner:
   builtinTunerName: Random
-  classArgs:
-    optimize_mode: maximize
 
 trial:
   codeDir: .

test/tuner_test.py

@@ -38,7 +38,7 @@ def switch(dispatch_type, dispatch_name):
     '''Change dispatch in config.yml'''
     config_path = 'tuner_test/local.yml'
     experiment_config = get_yml_content(config_path)
-    if dispatch_name in ['GridSearch', 'BatchTuner']:
+    if dispatch_name in ['GridSearch', 'BatchTuner', 'Random']:
         experiment_config[dispatch_type.lower()] = {
             'builtin' + dispatch_type + 'Name': dispatch_name
         }

tools/nni_cmd/config_schema.py

@@ -54,14 +54,14 @@ Optional('advisor'): Or({
     Optional('gpuNum'): And(int, lambda x: 0 <= x <= 99999),
 }),
 Optional('tuner'): Or({
-    'builtinTunerName': Or('TPE', 'Random', 'Anneal', 'SMAC', 'Evolution'),
+    'builtinTunerName': Or('TPE', 'Anneal', 'SMAC', 'Evolution'),
     Optional('classArgs'): {
         'optimize_mode': Or('maximize', 'minimize')
     },
     Optional('includeIntermediateResults'): bool,
     Optional('gpuNum'): And(int, lambda x: 0 <= x <= 99999),
 },{
-    'builtinTunerName': Or('BatchTuner', 'GridSearch'),
+    'builtinTunerName': Or('BatchTuner', 'GridSearch', 'Random'),
     Optional('gpuNum'): And(int, lambda x: 0 <= x <= 99999),
 },{
     'builtinTunerName': 'NetworkMorphism',