Extremely low accuracy on UCF with pre-trained model

[chengengliu]

Hi there,
thanks for the work. We are trying to reproduce the experiments, however, both training and validation top1 accuracy were closed to 0. We followed the configuration in another issue #3 , with batch_size adjusted:
python main.py --root_path ~/ --video_path ~/datasets/jester --annotation_path Efficient-3DCNNs/annotation_UCF101/ucf101_01.json --result_path Efficient-3DCNNs/results --pretrain_path Efficient-3DCNNs/results/kinetics_shufflenet_0.5x_G3_RGB_16_best.pth --dataset ucf101 --n_classes 600 --n_finetune_classes 101 --ft_portion last_layer --model shufflenet --groups 3 --width_mult 0.5 --train_crop random --learning_rate 0.1 --sample_duration 16 --batch_size 64 --n_threads 16 --checkpoint 1 --n_val_samples 1 \ Anyone could provide any thoughts? Thanks