fix bugs and update README

.gitignore

@@ -1,4 +1,5 @@
 .idea
+.git
 __pycache__
 run_all.sh
 get_time_all.sh

README.md

@@ -1,10 +1,12 @@
 # Multinomial Distribution Learning for Effective Neural Architecture Search
 
-Here we propose a method to extremely accelerate NAS, **without reinforcement learning or gradient**, just by sampling architectures from a distribution and comparing these architectures, Iteratively updating parameters of distribution while training
+Here we propose a method to **extremely accelerate** NAS, **without reinforcement learning or gradient**, just by sampling architectures from a distribution and comparing these architectures, estimating their **relative performance** rather than absolute performance, iteratively updating parameters of the distribution while training.
 
 ![](figs/1.png)
 
-Here we provide our test codes and pretrained model, our code is based on [DARTS](<https://github.com/khanrc/pt.darts>) and [ProxylessNAS](<https://github.com/mit-han-lab/ProxylessNAS>), pretrained models can be downloaded [here](https://drive.google.com/open?id=1W0UqwAnm37uibTuPDrH5Mt8PKNvFdD3v)
+Here we provide our test codes and pretrained models, our code is based on [DARTS](<https://github.com/khanrc/pt.darts>) and [ProxylessNAS](<https://github.com/mit-han-lab/ProxylessNAS>), pretrained models can be downloaded [here](https://drive.google.com/open?id=1W0UqwAnm37uibTuPDrH5Mt8PKNvFdD3v)
+
+**Search codes** will be released by [Sherwood](https://github.com/zhengxiawu) later !
 
 ## Requirements
 
@@ -13,7 +15,7 @@ Here we provide our test codes and pretrained model, our code is based on [DARTS
 
 ## Evaluate
 
-You need to modified your path to dataset in ``` data_providers/cifar10.py``` and ```data_providers/imagenet.py```
+You need to modified your path to dataset in ``` data_providers/cifar10.py``` and ```data_providers/imagenet.py```，```config.sh``` is used to prepare your environment, you should **write this file by yourself** and here we use it to prepare dataset and packages
 
 To evaluate the model in **DARTS setting**, just run
 

models/darts_nets_cifar/augment_cells.py

@@ -1,8 +1,8 @@
 """ CNN cell for network augmentation """
 import torch
 import torch.nn as nn
-from models.darts_nets_cifar import ops
 from utils import *
+from models.darts_nets_cifar import ops
 
 def to_dag(C_in, gene, reduction):
     """ generate discrete ops from gene """

models/darts_nets_imagenet/augment_cells.py

@@ -1,9 +1,28 @@
 """ CNN cell for network augmentation """
 import torch
 import torch.nn as nn
-from models.darts_nets_imagenet import ops
 from utils import *
+from models.darts_nets_imagenet import ops
 
+def to_dag(C_in, gene, reduction):
+    """ generate discrete ops from gene """
+    dag = nn.ModuleList()
+    for edges in gene:
+        row = nn.ModuleList()
+        for op_name, s_idx in edges:
+            # reduction cell & from input nodes => stride = 2
+            stride = 2 if reduction and s_idx < 2 else 1
+            op = ops.OPS[op_name](C_in, stride, True)
+            if not isinstance(op, ops.Identity):  # Identity does not use drop path
+                op = nn.Sequential(
+                    op,
+                    ops.DropPath_()
+                )
+            op.s_idx = s_idx
+            row.append(op)
+        dag.append(row)
+
+    return dag
 
 class AugmentCell(nn.Module):
     """ Cell for augmentation

models/darts_nets_imagenet/ops.py

@@ -186,24 +186,4 @@ def forward(self, x):
         return out
 
 
-class MixedOp(nn.Module):
-    """ Mixed operation """
 
-    def __init__(self, C, stride):
-        super().__init__()
-        self._ops = nn.ModuleList()
-        for primitive in PRIMITIVES:
-            # 利用索引来使用函数可以使用字典，键值分别为索引和lambda表达式
-            op = OPS[primitive](C, stride, affine=False)
-            self._ops.append(op)
-
-    def forward(self, x, weights):
-        """
-        Args:
-            x: input
-            weights: weight for each operation
-        """
-        return sum(w * op(x) for w, op in zip(weights, self._ops))
-        # index = torch.multinomial(weights, 1)
-        # sum = self._ops[index](x)
-        # return sum