Can't export the model to int8

Accelerated Computing Intelligent Video Analytics TAO Toolkit
1

Hi! Trying to convert yolov3_resnet18 to int8 with the following command:

!tlt-export yolo -m …/…/…/data/yolo_paycon_resnet18/experiment_dir_retrain_01/weights/yolo_resnet18_epoch_70.tlt
-o $USER_EXPERIMENT_DIR/export/yolo_paycon_resnet18_version2_bs_1_res_320_th_01_int8.etlt
-e $SPECS_DIR/yolo_paycon_retrain_resnet18_kitti.txt
-k $KEY
–cal_image_dir /workspace/tlt-experiments/data_paycon/training/kitti_train_images_version2/
–data_type int8
–batch_size 1
–batches 100
–cal_cache_file $USER_EXPERIMENT_DIR/export/cal.bin
–cal_data_file $USER_EXPERIMENT_DIR/export/cal.tensorfile

However, I get the following error: “ValueError: Cannot find input file name”. No clue what’s this about, could find any similar issues. Could you give me a hand?

3

Make sure your tlt file is available.

4

Found the error. The part was correct, but the KEY wasn’t. Again. Man, I hate these KEYs so much. Can’t imagine anything more pointless and useless than these KEYs. How difficult is it to add warnings or exceptions, if you don’t specify them (or use default), so that people don’t lose the training time many steps later.