Error Feature Generation

[ttmgr]

Hi again,

after fixing the path issue I am getting a different error by now:

Traceback (most recent call last):
File "main.py", line 368, in
main()
File "main.py", line 288, in main
train_CLmodel(logger,args)
File "/COMEBin/COMEBin/train_CLmodel.py", line 235, in train_CLmodel
simclr.train_addpretrain(train_loader, dataset, namelist)
File "/COMEBin/COMEBin/simclr.py", line 299, in train_addpretrain
top1, top5 = accuracy(logits, labels, topk=(1, 5))
UnboundLocalError: local variable 'logits' referenced before assignment
Something went wrong with running training network. Exiting.

Is there a way to fix this issue?

Best,
tim

[ziyewang]

Hi,

The issue may be due to the default parameters not being suitable for datasets with fewer than 1024 contigs. You can retry by adding this option when running run_comebin.sh:
"-b INT batch size for training process (default=1024)" and the numerical value can be set to the actual number of contigs in the specific dataset.

Best,
Ziye

[ziyewang]

Hi,

Besides adjusting the parameters, updating to the latest version (v1.0.4) can also resolve this issue.

Best,
Ziye

[ttmgr]

Sorry for the late response. It worked now, but i have another issue now :D