You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Hi! I really appreciate your fantastic work and code. And I've reproduced your work through the guidance in README.md
However, I've received this error when executing the training process with train.py.
The environments are based on the given environment.yml file, the version of torch are listed below: - OS: CentOS Linux release 8.4.2105 - Package Version:
python=3.7.10
pytorch=1.10.0=py3.7_cpu_0
torchaudio=0.10.0=py37_cpu
torchvision=0.11.1=py37_cpu
pytorch-cluster=1.5.9=py37_torch_1.10.0_cpu
pytorch-mutex=1.0=cpu
pytorch-scatter=2.0.9=py37_torch_1.10.0_cpu
pytorch-sparse=0.6.12=py37_torch_1.10.0_cpu
pytorch-spline-conv=1.2.1=py37_torch_1.10.0_cpu
torch-geometric=2.0.2
Additional context:
This error was raised while dataloader enumeration was called during training, i.e. for i, data in tqdm(enumerate(loader), total=len(loader)):. The Expected size 19 but got size 21 error during torch.cat comes from that it tried to cat tensor B (2nd molecule) with shape 10x21x3 to tensor A (1st molecule) with shape 10x19x3 at dimension 0 (10), which needs that the other dimension (19/21) should be the same. I'm not sure if this occurrence is normal to you and not sure where to make the modifications (if needed).
Looking forward to your reply :)
The text was updated successfully, but these errors were encountered:
Hmm I'm not sure immediately what the issue is, but I have a few suggestions. First, could you try downgrading torch-geometric to 1.6.3? I think that's the primary difference between my local versions and the versions you have listed.
Hmm I'm not sure immediately what the issue is, but I have a few suggestions. First, could you try downgrading torch-geometric to 1.6.3? I think that's the primary difference between my local versions and the versions you have listed.
It solves my issue. Thanks for your reply.
To be noted, pytorch 1.10.0 has some problems with torch-geometric 1.6.3. So I also downgraded pytorch to 1.7.0.
Hi! I really appreciate your fantastic work and code. And I've reproduced your work through the guidance in
README.md
However, I've received this error when executing the training process with
train.py
.Describe the error
To Reproduce
`python train.py --data_dir data/QM9/qm9/ --split_path data/QM9/splits/split0.npy --log_dir ./test_run --n_epochs 250 --dataset qm9`
Expected behavior
Training completed smoothly without error
Environments:
The environments are based on the given
environment.yml
file, the version of torch are listed below:- OS: CentOS Linux release 8.4.2105
- Package Version:
Additional context:
This error was raised while dataloader enumeration was called during training, i.e.
for i, data in tqdm(enumerate(loader), total=len(loader)):
. TheExpected size 19 but got size 21
error during torch.cat comes from that it tried to cat tensor B (2nd molecule) with shape10x21x3
to tensor A (1st molecule) with shape10x19x3
at dimension 0 (10), which needs that the other dimension (19/21) should be the same. I'm not sure if this occurrence is normal to you and not sure where to make the modifications (if needed).Looking forward to your reply :)
The text was updated successfully, but these errors were encountered: