Created by: hshen14

I would like to try training-aware INT8 for MobileNetSSD. I followed the instructions in https://github.com/PaddlePaddle/models/blob/develop/fluid/PaddleCV/object_detection/README_quant.md and ran the below command:

python main_quant.py --data_dir=/lustre/dataset/VOC/VOCdevkit/VOC2012 --mode='train' --init_model=ssd_mobilenet_v1_pascalvoc --act_quant_type='abs_max' --epoc_num=20 --learning_rate=0.0001 --batch_size=64 --model_save_dir=train_int8

It shows the error message: Traceback (most recent call last): File "main_quant.py", line 275, in ap_version = args.ap_version) File "/home/hshen14/paddle_public_models/fluid/PaddleCV/object_detection/reader.py", line 45, in init for line in open(label_fpath): IOError: [Errno 2] No such file or directory: '/lustre/dataset/VOC/VOCdevkit/VOC2012/label_list'

I attached the structure under /lustre/dataset/VOC/VOCdevkit/VOC2012/ for your reference: Annotations/ JPEGImages/ SegmentationObject/ ImageSets/ SegmentationClass/

From the error message, looks like there is no label_list file. Could you please show the full structure at you side? Thanks.

@qingqing01 @Superjomn