Support GDAS (FRC), see details in docs/CVPR-2019-GDAS.md

configs/search-archs/GDASFRC-NASNet-CIFAR.config

@@ -0,0 +1,9 @@
+{
+  "super_type"      : ["str",  "nasnet-super"],
+  "name"            : ["str",  "GDAS_FRC"],
+  "C"               : ["int",  "16"  ],
+  "N"               : ["int",  "2"  ],
+  "steps"           : ["int",  "4"  ],
+  "multiplier"      : ["int",  "4"  ],
+  "stem_multiplier" : ["int",  "3"  ]
+}

docs/CVPR-2019-GDAS.md

@@ -37,9 +37,14 @@ CUDA_VISIBLE_DEVICES=0,1,2,3 bash ./scripts/nas-infer-train.sh imagenet-1k GDAS_
 If you are interested in the configs of each NAS-searched architecture, they are defined at [genotypes.py](https://github.com/D-X-Y/AutoDL-Projects/blob/master/lib/nas_infer_model/DXYs/genotypes.py).
 
 ### Searching on the NASNet search space
+
 Please use the following scripts to use GDAS to search as in the original paper:
 ```
+# search for both normal and reduction cells
 CUDA_VISIBLE_DEVICES=0 bash ./scripts-search/NASNet-space-search-by-GDAS.sh cifar10 1 -1
+
+# search for the normal cell while use a fixed reduction cell
+CUDA_VISIBLE_DEVICES=0 bash ./scripts-search/NASNet-space-search-by-GDAS-FRC.sh cifar10 1 -1
 ```
 
 **After searching**, if you want to re-train the searched architecture found by the above script, you can use the following script:
@@ -52,7 +57,9 @@ Note that `gdas-searched` is a string to indicate the name of the saved dir and
 The above script does not apply heavy augmentation to train the model, so the accuracy will be lower than the original paper.
 If you want to change the default hyper-parameter for re-training, please have a look at `./scripts/retrain-searched-net.sh` and `configs/archs/NAS-*-none.config`.
 
+
 ### Searching on a small search space (NAS-Bench-201)
+
 The GDAS searching codes on a small search space:
 ```
 CUDA_VISIBLE_DEVICES=0 bash ./scripts-search/algos/GDAS.sh cifar10 1 -1

lib/models/cell_operations.py

@@ -4,7 +4,7 @@
 import torch
 import torch.nn as nn
 
-__all__ = ['OPS', 'ResNetBasicblock', 'SearchSpaceNames']
+__all__ = ['OPS', 'RAW_OP_CLASSES', 'ResNetBasicblock', 'SearchSpaceNames']
 
 OPS = {
   'none'         : lambda C_in, C_out, stride, affine, track_running_stats: Zero(C_in, C_out, stride),
@@ -175,7 +175,7 @@ class FactorizedReduce(nn.Module):
         self.convs.append(nn.Conv2d(C_in, C_outs[i], 1, stride=stride, padding=0, bias=not affine))
       self.pad = nn.ConstantPad2d((0, 1, 0, 1), 0)
     elif stride == 1:
-      self.conv = nn.Conv2d(C_in, C_out, 1, stride=stride, padding=0, bias=False)
+      self.conv = nn.Conv2d(C_in, C_out, 1, stride=stride, padding=0, bias=not affine)
     else:
       raise ValueError('Invalid stride : {:}'.format(stride))
     self.bn = nn.BatchNorm2d(C_out, affine=affine, track_running_stats=track_running_stats)
@@ -256,41 +256,44 @@ def drop_path(x, drop_prob):
 # Searching for A Robust Neural Architecture in Four GPU Hours
 class GDAS_Reduction_Cell(nn.Module):
 
-  def __init__(self, C_prev_prev, C_prev, C, reduction_prev, multiplier, affine, track_running_stats):
+  def __init__(self, C_prev_prev, C_prev, C, reduction_prev, affine, track_running_stats):
     super(GDAS_Reduction_Cell, self).__init__()
     if reduction_prev:
       self.preprocess0 = FactorizedReduce(C_prev_prev, C, 2, affine, track_running_stats)
     else:
       self.preprocess0 = ReLUConvBN(C_prev_prev, C, 1, 1, 0, 1, affine, track_running_stats)
     self.preprocess1 = ReLUConvBN(C_prev, C, 1, 1, 0, 1, affine, track_running_stats)
-    self.multiplier  = multiplier
 
     self.reduction = True
     self.ops1 = nn.ModuleList(
                   [nn.Sequential(
                       nn.ReLU(inplace=False),
-                      nn.Conv2d(C, C, (1, 3), stride=(1, 2), padding=(0, 1), groups=8, bias=False),
-                      nn.Conv2d(C, C, (3, 1), stride=(2, 1), padding=(1, 0), groups=8, bias=False),
-                      nn.BatchNorm2d(C, affine=True),
+                      nn.Conv2d(C, C, (1, 3), stride=(1, 2), padding=(0, 1), groups=8, bias=not affine),
+                      nn.Conv2d(C, C, (3, 1), stride=(2, 1), padding=(1, 0), groups=8, bias=not affine),
+                      nn.BatchNorm2d(C, affine=affine, track_running_stats=track_running_stats),
                       nn.ReLU(inplace=False),
-                      nn.Conv2d(C, C, 1, stride=1, padding=0, bias=False),
-                      nn.BatchNorm2d(C, affine=True)),
+                      nn.Conv2d(C, C, 1, stride=1, padding=0, bias=not affine),
+                      nn.BatchNorm2d(C, affine=affine, track_running_stats=track_running_stats)),
                    nn.Sequential(
                       nn.ReLU(inplace=False),
-                      nn.Conv2d(C, C, (1, 3), stride=(1, 2), padding=(0, 1), groups=8, bias=False),
-                      nn.Conv2d(C, C, (3, 1), stride=(2, 1), padding=(1, 0), groups=8, bias=False),
-                      nn.BatchNorm2d(C, affine=True),
+                      nn.Conv2d(C, C, (1, 3), stride=(1, 2), padding=(0, 1), groups=8, bias=not affine),
+                      nn.Conv2d(C, C, (3, 1), stride=(2, 1), padding=(1, 0), groups=8, bias=not affine),
+                      nn.BatchNorm2d(C, affine=affine, track_running_stats=track_running_stats),
                       nn.ReLU(inplace=False),
-                      nn.Conv2d(C, C, 1, stride=1, padding=0, bias=False),
-                      nn.BatchNorm2d(C, affine=True))])
+                      nn.Conv2d(C, C, 1, stride=1, padding=0, bias=not affine),
+                      nn.BatchNorm2d(C, affine=affine, track_running_stats=track_running_stats))])
 
     self.ops2 = nn.ModuleList(
                   [nn.Sequential(
-                      nn.MaxPool2d(3, stride=1, padding=1),
-                      nn.BatchNorm2d(C, affine=True)),
+                      nn.MaxPool2d(3, stride=2, padding=1),
+                      nn.BatchNorm2d(C, affine=affine, track_running_stats=track_running_stats)),
                    nn.Sequential(
                       nn.MaxPool2d(3, stride=2, padding=1),
-                      nn.BatchNorm2d(C, affine=True))])
+                      nn.BatchNorm2d(C, affine=affine, track_running_stats=track_running_stats))])
+
+  @property
+  def multiplier(self):
+    return 4
 
   def forward(self, s0, s1, drop_prob = -1):
     s0 = self.preprocess0(s0)
@@ -307,3 +310,10 @@ class GDAS_Reduction_Cell(nn.Module):
     if self.training and drop_prob > 0.:
       X2, X3 = drop_path(X2, drop_prob), drop_path(X3, drop_prob)
     return torch.cat([X0, X1, X2, X3], dim=1)
+
+
+# To manage the useful classes in this file.
+RAW_OP_CLASSES = {
+  'gdas_reduction': GDAS_Reduction_Cell
+}
+

lib/models/cell_searchs/__init__.py

@@ -11,6 +11,7 @@ from .generic_model         import GenericNAS201Model
 from .genotypes             import Structure as CellStructure, architectures as CellArchitectures
 # NASNet-based macro structure
 from .search_model_gdas_nasnet import NASNetworkGDAS
+from .search_model_gdas_frc_nasnet import NASNetworkGDAS_FRC
 from .search_model_darts_nasnet import NASNetworkDARTS
 
 
@@ -23,4 +24,5 @@ nas201_super_nets = {'DARTS-V1': TinyNetworkDarts,
                      "generic": GenericNAS201Model}
 
 nasnet_super_nets = {"GDAS": NASNetworkGDAS,
+                     "GDAS_FRC": NASNetworkGDAS_FRC,
                      "DARTS": NASNetworkDARTS}

lib/models/cell_searchs/search_cells.py

@@ -163,6 +163,10 @@ class NASNetSearchCell(nn.Module):
     self.edge2index = {key:i for i, key in enumerate(self.edge_keys)}
     self.num_edges  = len(self.edges)
 
+  @property
+  def multiplier(self):
+    return self._multiplier
+
   def forward_gdas(self, s0, s1, weightss, indexs):
     s0 = self.preprocess0(s0)
     s1 = self.preprocess1(s1)

lib/models/cell_searchs/search_model_gdas_frc_nasnet.py

@@ -0,0 +1,125 @@
+###########################################################################
+# Searching for A Robust Neural Architecture in Four GPU Hours, CVPR 2019 #
+###########################################################################
+import torch
+import torch.nn as nn
+from copy import deepcopy
+from models.cell_searchs.search_cells import NASNetSearchCell as SearchCell
+from models.cell_operations import RAW_OP_CLASSES
+
+
+# The macro structure is based on NASNet
+class NASNetworkGDAS_FRC(nn.Module):
+
+  def __init__(self, C, N, steps, multiplier, stem_multiplier, num_classes, search_space, affine, track_running_stats):
+    super(NASNetworkGDAS_FRC, self).__init__()
+    self._C        = C
+    self._layerN   = N
+    self._steps    = steps
+    self._multiplier = multiplier
+    self.stem = nn.Sequential(
+                    nn.Conv2d(3, C*stem_multiplier, kernel_size=3, padding=1, bias=False),
+                    nn.BatchNorm2d(C*stem_multiplier))
+  
+    # config for each layer
+    layer_channels   = [C    ] * N + [C*2 ] + [C*2  ] * (N-1) + [C*4 ] + [C*4  ] * (N-1)
+    layer_reductions = [False] * N + [True] + [False] * (N-1) + [True] + [False] * (N-1)
+
+    num_edge, edge2index = None, None
+    C_prev_prev, C_prev, C_curr, reduction_prev = C*stem_multiplier, C*stem_multiplier, C, False
+
+    self.cells = nn.ModuleList()
+    for index, (C_curr, reduction) in enumerate(zip(layer_channels, layer_reductions)):
+      if reduction:
+        cell = RAW_OP_CLASSES['gdas_reduction'](C_prev_prev, C_prev, C_curr, reduction_prev, affine, track_running_stats)
+      else:
+        cell = SearchCell(search_space, steps, multiplier, C_prev_prev, C_prev, C_curr, reduction, reduction_prev, affine, track_running_stats)
+      if num_edge is None: num_edge, edge2index = cell.num_edges, cell.edge2index
+      else: assert reduction or num_edge == cell.num_edges and edge2index == cell.edge2index, 'invalid {:} vs. {:}.'.format(num_edge, cell.num_edges)
+      self.cells.append( cell )
+      C_prev_prev, C_prev, reduction_prev = C_prev, cell.multiplier * C_curr, reduction
+    self.op_names   = deepcopy( search_space )
+    self._Layer     = len(self.cells)
+    self.edge2index = edge2index
+    self.lastact    = nn.Sequential(nn.BatchNorm2d(C_prev), nn.ReLU(inplace=True))
+    self.global_pooling = nn.AdaptiveAvgPool2d(1)
+    self.classifier = nn.Linear(C_prev, num_classes)
+    self.arch_parameters = nn.Parameter( 1e-3*torch.randn(num_edge, len(search_space)) )
+    self.tau        = 10
+
+  def get_weights(self):
+    xlist = list( self.stem.parameters() ) + list( self.cells.parameters() )
+    xlist+= list( self.lastact.parameters() ) + list( self.global_pooling.parameters() )
+    xlist+= list( self.classifier.parameters() )
+    return xlist
+
+  def set_tau(self, tau):
+    self.tau = tau
+
+  def get_tau(self):
+    return self.tau
+
+  def get_alphas(self):
+    return [self.arch_parameters]
+
+  def show_alphas(self):
+    with torch.no_grad():
+      A = 'arch-normal-parameters :\n{:}'.format(nn.functional.softmax(self.arch_parameters, dim=-1).cpu())
+    return '{:}'.format(A)
+
+  def get_message(self):
+    string = self.extra_repr()
+    for i, cell in enumerate(self.cells):
+      string += '\n {:02d}/{:02d} :: {:}'.format(i, len(self.cells), cell.extra_repr())
+    return string
+
+  def extra_repr(self):
+    return ('{name}(C={_C}, N={_layerN}, steps={_steps}, multiplier={_multiplier}, L={_Layer})'.format(name=self.__class__.__name__, **self.__dict__))
+
+  def genotype(self):
+    def _parse(weights):
+      gene = []
+      for i in range(self._steps):
+        edges = []
+        for j in range(2+i):
+          node_str = '{:}<-{:}'.format(i, j)
+          ws = weights[ self.edge2index[node_str] ]
+          for k, op_name in enumerate(self.op_names):
+            if op_name == 'none': continue
+            edges.append( (op_name, j, ws[k]) )
+        edges = sorted(edges, key=lambda x: -x[-1])
+        selected_edges = edges[:2]
+        gene.append( tuple(selected_edges) )
+      return gene
+    with torch.no_grad():
+      gene_normal = _parse(torch.softmax(self.arch_parameters, dim=-1).cpu().numpy())
+    return {'normal': gene_normal, 'normal_concat': list(range(2+self._steps-self._multiplier, self._steps+2))}
+
+  def forward(self, inputs):
+    def get_gumbel_prob(xins):
+      while True:
+        gumbels = -torch.empty_like(xins).exponential_().log()
+        logits  = (xins.log_softmax(dim=1) + gumbels) / self.tau
+        probs   = nn.functional.softmax(logits, dim=1)
+        index   = probs.max(-1, keepdim=True)[1]
+        one_h   = torch.zeros_like(logits).scatter_(-1, index, 1.0)
+        hardwts = one_h - probs.detach() + probs
+        if (torch.isinf(gumbels).any()) or (torch.isinf(probs).any()) or (torch.isnan(probs).any()):
+          continue
+        else: break
+      return hardwts, index
+
+    hardwts, index = get_gumbel_prob(self.arch_parameters)
+
+    s0 = s1 = self.stem(inputs)
+    for i, cell in enumerate(self.cells):
+      if cell.reduction:
+        s0, s1 = s1, cell(s0, s1)
+      else: 
+        s0, s1 = s1, cell.forward_gdas(s0, s1, hardwts, index)
+    out = self.lastact(s1)
+    out = self.global_pooling( out )
+    out = out.view(out.size(0), -1)
+    logits = self.classifier(out)
+
+    return out, logits

scripts-search/NASNet-space-search-by-GDAS-FRC.sh

@@ -0,0 +1,38 @@
+#!/bin/bash
+# bash ./scripts-search/NASNet-space-search-by-GDAS-FRC.sh cifar10 1 -1
+echo script name: $0
+echo $# arguments
+if [ "$#" -ne 3 ] ;then
+  echo "Input illegal number of parameters " $#
+  echo "Need 3 parameters for dataset, track_running_stats, and seed"
+  exit 1
+fi
+if [ "$TORCH_HOME" = "" ]; then
+  echo "Must set TORCH_HOME envoriment variable for data dir saving"
+  exit 1
+else
+  echo "TORCH_HOME : $TORCH_HOME"
+fi
+
+dataset=$1
+track_running_stats=$2
+seed=$3
+space=darts
+
+if [ "$dataset" == "cifar10" ] || [ "$dataset" == "cifar100" ]; then
+  data_path="$TORCH_HOME/cifar.python"
+else
+  data_path="$TORCH_HOME/cifar.python/ImageNet16"
+fi
+
+save_dir=./output/search-cell-${space}/GDAS-${dataset}-BN${track_running_stats}
+
+OMP_NUM_THREADS=4 python ./exps/algos/GDAS.py \
+	--save_dir ${save_dir} \
+	--dataset ${dataset} --data_path ${data_path} \
+	--search_space_name ${space} \
+	--config_path  configs/search-opts/GDAS-NASNet-CIFAR.config \
+	--model_config configs/search-archs/GDASFRC-NASNet-CIFAR.config \
+	--tau_max 10 --tau_min 0.1 --track_running_stats ${track_running_stats} \
+	--arch_learning_rate 0.0003 --arch_weight_decay 0.001 \
+	--workers 4 --print_freq 200 --rand_seed ${seed}