How to set a default evaluator using the configuration file

[Nasrine-b]

Hi!
I have a dataset of resumes (train - validation). I want to re-train a "Faster-RCNN-R-50-FPN-3x" for object detection on resumes (actually it's for document layout analysis).
So I ran my code on AWS. And I had a problem with the validation step.
Here's the config file :

cfg = get_cfg()
cfg.merge_from_file(model_zoo.get_config_file("COCO-Detection/faster_rcnn_R_50_FPN_3x.yaml"))#"COCO-InstanceSegmentation/mask_rcnn_R_50_FPN_3x.yaml")) #Get the basic model configuration from the model zoo
#Passing the Train and Validation sets
cfg.DATASETS.TRAIN = (nameTrainDS,)
cfg.DATASETS.TEST = (nameValDS,)
cfg.OUTPUT_DIR = os.environ["SM_MODEL_DIR"]
# Number of data loading threads
cfg.DATALOADER.NUM_WORKERS = 1
cfg.MODEL.WEIGHTS = model_zoo.get_checkpoint_url("COCO-Detection/faster_rcnn_R_50_FPN_3x.yaml")#"COCO-InstanceSegmentation/mask_rcnn_R_50_FPN_3x.yaml")#  # Let training initialize from model zoo
# Number of images per batch across all machines.
cfg.SOLVER.IMS_PER_BATCH = 2
cfg.SOLVER.BASE_LR = 0.0005  # pick a good LearningRate
cfg.SOLVER.MAX_ITER = 80  #No. of iterations
cfg.SOLVER.CHECKPOINT_PERIOD = 10 #Save a checkpoint after every this number of iterations
cfg.SOLVER.STEPS = []
cfg.MODEL.ROI_HEADS.BATCH_SIZE_PER_IMAGE = 128
cfg.MODEL.DEVICE = "cuda"
cfg.MODEL.ROI_HEADS.NUM_CLASSES = 2 # No. of classes = [Block, Subblock]
cfg.TEST.EVAL_PERIOD = 80 # No. of iterations after which the Validation Set is evaluated.
os.makedirs(cfg.OUTPUT_DIR, exist_ok=True)

The training phase ran without any error. But, as you can see on my configuration file, I have an evaluation at the end of the train, and I had the following error :
#033[5m#033[31mWARNING#033[0m #033[32m[07/20 13:26:09 d2.engine.defaults]: #033[0mNo evaluator found. Use `DefaultTrainer.test(evaluators=)`, or implement its `build_evaluator` method.

I looked on `DefaultTrainer class of Detectron2 and the default evaluator is set to None.
Based on Detectron2 documentation, I can choose between COCOEvaluator and SemSegEvaluator (because I used a custom dataset).
But I don't know how to call the evaluator using the config file. I'm planning on changing the evaluation period to 10 iterations, that's why I really want to call it on the config and not after the train.

Does someone have any idea to fix this issue ?
(Sorry for the english level btw)