Python train.py -h Validation and the real-time implementation Python train.py -name ANC_LSTM -continue_train -checkpoint_epoch XXXįor more details about the settings of training, To ine-tune the pretrained model at XXX iteration,
Python train.py -name vanilla_LSTM -att_dim 0 To train the ANC-LSTM model for the first time use the following command under the root path of the repo.įor the vanilla-LSTM model without ANC module for the comparison experiment, The training set and validation set are randomly spilt and each sequential sample is formatted in json with the inputs collected from the real-time system and the corresponding labels. This repo has been well organized with dataset in dataset folder and pretrained models in outputs folder, where the experimental results could be easily reproduced or extended for further research. The source code, pretrained models and dataset are released here for our IROS 2021 paper of "Soft Manipulator Fault Detection and Identification Using ANC-based LSTM" by Haoyuan Gu †, Hanjiang Hu †, Hesheng Wang* and Weidong Chen.
0 kommentar(er)
