increase l2 to 1e-4

bupticybee · Oct 12, 2018 · d89d908 · d89d908
1 parent 8163b47
commit d89d908
Show file tree

Hide file tree

Showing 4 changed files with 309 additions and 214 deletions.
diff --git a/ipynbs/elo_graph.ipynb b/ipynbs/elo_graph.ipynb
diff --git a/ipynbs/see_gameplay.ipynb b/ipynbs/see_gameplay.ipynb
diff --git a/ipynbs/vs_human.ipynb b/ipynbs/vs_human.ipynb
diff --git a/net/resnet.py b/net/resnet.py
@@ -180,7 +180,7 @@ def get_model(MODEL_NAME,labels,GPU_CORE = [0],BATCH_SIZE = 512,NUM_RES_LAYERS =
                             labels=score[ind * (BATCH_SIZE // len(GPU_CORE)):(ind + 1) * (BATCH_SIZE // len(GPU_CORE))],
                             predictions=value_head) 
                         value_loss = tf.reduce_mean(value_loss)
-                        regularizer = tf.contrib.layers.l2_regularizer(scale=1e-6)
+                        regularizer = tf.contrib.layers.l2_regularizer(scale=1e-4)
                         regular_variables = tf.trainable_variables()
                         l2_loss = tf.contrib.layers.apply_regularization(regularizer, regular_variables)
                         multitarget_loss = value_loss + policy_loss + l2_loss