NaN in Mask R-CNN training on MS COCO

It only happens to MS COCO, not Pascal VOC or ICDAR. I fixed the class ids:

coco_interface = COCO("instances_train2017.json")
all_cats = coco_interface.getCatIds()
all_cats.insert(0,0)
print(all_cats, len(all_cats))

[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 27, 28, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 67, 70, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 84, 85, 86, 87, 88, 89, 90]

selected_class_ids = coco_interface.getCatIds(catNms=['person'])
adjusted_class_ids = {}
for id, cl in enumerate(all_cats):
    adjusted_class_ids[cl] = id
print ("ADJUSTED CLASS IDS:")
print(adjusted_class_ids)

{0: 0, 1: 1, 2: 2, 3: 3, 4: 4, 5: 5, 6: 6, 7: 7, 8: 8, 9: 9, 10: 10, 11: 11, 13: 12, 14: 13, 15: 14, 16: 15, 17: 16, 18: 17, 19: 18, 20: 19, 21: 20, 22: 21, 23: 22, 24: 23, 25: 24, 27:25, 28: 26, 31: 27, 32: 28, 33: 29, 34: 30, 35: 31, 36: 32, 37: 33, 38: 34, 39: 35, 40: 36, 41: 37, 42: 38, 43: 39, 44: 40, 46: 41, 47: 42, 48: 43, 49: 44, 50: 45, 51: 46, 52…88: 78, 89: 79, 90: 80}

The class label is converted so that it’s between 0 and 80. The total number of classes is 81. Learning rate is very low. What else can I do?