When I trained on VOC2007:
Create YOLOv3 model with 9 anchors and 20 classes.
Load weights /home/gpu/projects/XS/keras-zero-shot-detection-master/data/darknet53_weights.h5.
Freeze the first 249 layers of total 252 layers.
Train on 8967 samples, val on 996 samples, with batch size 8.
Epoch 1/10
1/1000 [..............................] - ETA: 6:02:42 - loss: nan
2/1000 [..............................] - ETA: 3:50:33 - loss: nan
3/1000 [..............................] - ETA: 3:10:26 - loss: nan
4/1000 [..............................] - ETA: 2:49:20 - loss: nan
When I trained on VOC2007: Create YOLOv3 model with 9 anchors and 20 classes. Load weights /home/gpu/projects/XS/keras-zero-shot-detection-master/data/darknet53_weights.h5. Freeze the first 249 layers of total 252 layers. Train on 8967 samples, val on 996 samples, with batch size 8. Epoch 1/10
1/1000 [..............................] - ETA: 6:02:42 - loss: nan 2/1000 [..............................] - ETA: 3:50:33 - loss: nan 3/1000 [..............................] - ETA: 3:10:26 - loss: nan 4/1000 [..............................] - ETA: 2:49:20 - loss: nan
python:3.5 keras:2.2 tensorflow:1.10
what's wrong with me?