devopsgitpython
diff --git a/‎homework03/homework03_part1.ipynb
Lines changed: 174 additions & 0 deletions b/‎homework03/homework03_part1.ipynb
Lines changed: 174 additions & 0 deletions
diff --git a/‎week05_dense_prediction/README.md
Lines changed: 9 additions & 0 deletions b/‎week05_dense_prediction/README.md
Lines changed: 9 additions & 0 deletions
diff --git a/‎week05_dense_prediction/layer_wrappers.py
Lines changed: 46 additions & 0 deletions b/‎week05_dense_prediction/layer_wrappers.py
Lines changed: 46 additions & 0 deletions
diff --git a/‎week05_dense_prediction/seminar_dense_prediction.ipynb
Lines changed: 736 additions & 0 deletions b/‎week05_dense_prediction/seminar_dense_prediction.ipynb
Lines changed: 736 additions & 0 deletions
@@ -0,0 +1,9 @@
+
+## Practice
+
+[![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/yandexdataschool/Practical_DL/blob/spring2019/week05_dense_prediction/seminar_dense_prediction.ipynb)
+
+Open seminar_dense_prediction.ipynb and follow instructions from there.
+
+
+
@@ -0,0 +1,46 @@
+import torch
+import torch.nn as nn
+from torch.autograd import Variable
+
+
+def init_layer(layer, weight_init=None, bias_init=None):
+    if weight_init is not None:
+        layer.weight.data = torch.FloatTensor(weight_init)
+    if bias_init is not None:
+        layer.bias.data = torch.FloatTensor(bias_init)
+    return layer
+
+
+def init_bn_statistics(layer, mean_init=None, var_init=None):
+    if mean_init is not None:
+        layer.running_mean = torch.FloatTensor(mean_init)
+    if var_init is not None:
+        layer.running_var = torch.FloatTensor(var_init)
+    return layer
+
+
+def Linear(in_features, out_features, bias=True, weight_init=None, bias_init=None):
+    layer = nn.Linear(in_features, out_features, bias)
+    return init_layer(layer, weight_init, bias_init)
+
+
+def Conv2d(in_channels, out_channels, kernel_size, stride=1, padding=0, dilation=1, groups=1,
+           bias=True, weight_init=None, bias_init=None):
+    layer = nn.Conv2d(in_channels, out_channels, kernel_size, stride, padding, dilation,
+                      groups, bias)
+    return init_layer(layer, weight_init, bias_init)
+
+
+def BatchNorm1d(num_features, eps=1e-05, momentum=0.1, affine=True, weight_init=None,
+                bias_init=None, mean_init=None, var_init=None):
+    layer = nn.BatchNorm1d(num_features, eps, momentum, affine)
+    layer = init_layer(layer, weight_init, bias_init)
+    return init_bn_statistics(layer, mean_init, var_init)
+
+
+def BatchNorm2d(num_features, eps=1e-05, momentum=0.1, affine=True, weight_init=None,
+                bias_init=None, mean_init=None, var_init=None):
+    layer = nn.BatchNorm2d(num_features, eps, momentum, affine)
+    layer = init_layer(layer, weight_init, bias_init)
+    return init_bn_statistics(layer, mean_init, var_init)
+