cram3r95 / mapfe4mp Goto Github PK

Hi folks,
I followed your README and ran the default ``python train.py --trainer mapfe4mp --device_gpu 0"
The first time, error reports not found preprocessed data.

Then I modified config_mapfe4mp.yml to enable preprocess_data and save_data.

Then I run again but encountered the issue that missing
arget_agent_orientation.npy
oracle_centerlines.npy
relevant_centerlines.npy
during preprocessing.

This breaks data_utils.py function load_processed_files_from_npy, as required_variables_name_list not all satisfied.
I wonder how to perform preprocessing properly?
thank you.

May I use train again_ Gan_ Social_ LSTM_ The following errors occurred during the training of the MHSA model：
Traceback (most recent call last):
File "/mnt/f/pycharm-professional/pytorch-project/VectorNet-GAN/mapfe4mp-master/train.py", line 143, in
model_trainer(config, logger)
File "/mnt/f/pycharm-professional/pytorch-project/VectorNet-GAN/mapfe4mp-master/model/trainers/other/trainer_gan_social_lstm_mhsa.py", line 271, in model_trainer
losses_d = discriminator_step(hyperparameters, batch, generator,
File "/mnt/f/pycharm-professional/pytorch-project/VectorNet-GAN/mapfe4mp-master/model/trainers/other/trainer_gan_social_lstm_mhsa.py", line 489, in discriminator_step
pred_traj_fake = relative_to_abs(pred_traj_fake_rel, obs_traj[-1])
File "/mnt/f/pycharm-professional/pytorch-project/VectorNet-GAN/mapfe4mp-master/model/datasets/argoverse/dataset_utils.py", line 394, in relative_to_abs
rel_traj = rel_traj.permute(1, 0, 2)
AttributeError: 'tuple' object has no attribute 'permute'
Excuse me, class TrajectoryGenerator (nn. Module): Is this part of the code correct and why is it returning pre d_ Traj_ Fake_ Rel returns a tuple, but tuples cannot use relative_ To_ ABS function

Can you upload that 'tutorial notebook for generating these visualizations' that you mentioned in Qualitative results or any python file for generating gif file for argoverse v1 .