RuntimeError: unique_by_key: failed to synchronize: cudaErrorIllegalAddress: an illegal memory access was encountered

[XJL0930]

Has anyone else had this problem when run "python train_transe_FB15K237.py"?

[yierhoho]

I have encountered the same problem. If the code reports an error in the first epoch, then you may not have set the batch_size properly. Try this in train_transe_FB15K237.py

dataloader for training

train_dataloader = TrainDataLoader(
in_path = "./benchmarks/FB15K237/",
nbatches = 10,
threads = 8,
sampling_mode = "normal",
bern_flag = 1,
filter_flag = 1,
neg_ent = 25,
neg_rel = 0,
batch_size=16)