HarryShomer/KG-Mixup

Thanks for sharing your work, but the model source code seems to have a bug.

Closed this issue · 2 comments

Epoch 1: 0%| | 0/1169 [00:01<?, ?it/s]
Traceback (most recent call last):
File "src/main.py", line 186, in
main()
File "src/main.py", line 182, in main
run_model(model, optimizer, data)
File "src/main.py", line 169, in run_model
model_trainer.fit(args.epochs, synth_gen, **train_fit_keywords)
File "/home/admin1/code/lly/2023-4/Mix/KG-Mixup/src/training.py", line 122, in fit
loss = self._train_batch(batch, train_type, label_smooth)
File "/home/admin1/code/lly/2023-4/Mix/KG-Mixup/kgpy/kgpy/training.py", line 193, in _train_batch
batch_loss = self._train_batch_1_to_n(batch, label_smooth)
File "/home/admin1/code/lly/2023-4/Mix/KG-Mixup/kgpy/kgpy/training.py", line 271, in _train_batch_1_to_n
all_scores = self.model(trips, mode="tail")
File "/home/admin1/anaconda3/envs/torch10/lib/python3.7/site-packages/torch/nn/modules/module.py", line 1194, in _call_impl
return forward_call(*input, **kwargs)
File "/home/admin1/code/lly/2023-4/Mix/KG-Mixup/kgpy/kgpy/models/base/base_emb_model.py", line 239, in forward
scores = self.score_tail(triplets)
File "/home/admin1/code/lly/2023-4/Mix/KG-Mixup/kgpy/kgpy/models/conve.py", line 184, in score_tail
return self.score_head(triplets)
File "/home/admin1/code/lly/2023-4/Mix/KG-Mixup/kgpy/kgpy/models/conve.py", line 164, in score_head
x = torch.mm(x, self.ent_embs.weight.transpose(1,0))
RuntimeError: CUDA error: CUBLAS_STATUS_INVALID_VALUE when calling cublasSgemm( handle, opa, opb, m, n, k, &alpha, a, lda, b, ldb, &beta, c, ldc)

@2391134843

Thanks for bringing this to my attention.

Could you share the command you're running?

Sorry, it was my mistake that led to this problem, based on StarE and the author of your paper model. I have this problem every time I try it. The reason is that the Python version is not correct. Python version must be >3.7. The bug disappeared after I used 3.8 and 3.9.