search

valeoai / WaffleIron

Other
39 stars 6 forks source link

IndexError: tensors used as indices must be long, byte or bool tensors #11

Closed SpeedyGonzales949 closed 3 weeks ago

SpeedyGonzales949 commented 4 weeks ago

I am using python 3.10.14 and pytorch 1.11.0 with cudatoolkit 11.3.1. I am trying to train on a RTX 2080 TI using the following config:
{ "label": "Run Train on Nuscenes", "type": "process", "command": "\anaconda3\envs\WaffleIron\python.exe", "args": [ "launch_train.py", "--dataset", "nuscenes", "--path_dataset", ".\data", "--log_path", ".\logs\WaffleIron-48-256nuscenes\", "--config", ".\configs\WaffleIron-48-256nuscenes.yaml", "--fp16", "--multiprocessing-distributed", "--gpu", "0", "--restart" ] } I get the following error:

Traceback (most recent call last): File "WaffleIron\launch_train.py", line 372, in main(args, config) File "WaffleIron\launch_train.py", line 306, in main distributed_training(args.gpu, ngpus_per_node, args, config) File "WaffleIron\launch_train.py", line 253, in distributed_training mng.train() File "WaffleIron\utils\trainer.py", line 290, in train self.one_epoch(training=True) File "WaffleIron\utils\trainer.py", line 173, in one_epoch temp = out[id_b, :, closest_point] IndexError: tensors used as indices must be long, byte or bool tensors

gpuy commented 3 weeks ago

Thank you. I accepted your pull request.