Merge pull request #916 from will-jl944/develop_jf

set pretrain_weights to None in prune tutorial

dygraph/tutorials/slim/prune/object_detection/yolov3_prune.py

@@ -60,6 +60,7 @@ model.train(
     train_dataset=train_dataset,
     train_batch_size=8,
     eval_dataset=eval_dataset,
+    pretrain_weights=None,
     learning_rate=0.001 / 8,
     warmup_steps=1000,
     warmup_start_lr=0.0,

dygraph/tutorials/slim/prune/semantic_segmentation/unet_prune.py

@@ -56,5 +56,6 @@ model.train(
     train_dataset=train_dataset,
     train_batch_size=4,
     eval_dataset=eval_dataset,
+    pretrain_weights=None,
     learning_rate=0.01,
     save_dir='output/unet/prune')