nulog / nulog Goto Github PK

what is the "ground_truth" type in fine_tuning the Nulog log parser?
Thanks for your sharing

Nick work!

But I can not find the demo of NuLog in this repo.

So I was wondering if I can just bring the parameters from NuLog_benchmark.py as a demo of NuLog to parse the whole dataset?
Thanks

Thanks for sharing your code
How can I use this code for the testing step after training and load the saved model?
I write this:
model = torch.load(_path,'cuda')
but I get this error:
yield out_p.cpu().numpy().argsort(axis=1), b_labels.cpu().numpy(), ind.cpu()
RuntimeError: CUDA error: device-side assert triggered
/pytorch/aten/src/ATen/native/cuda/Indexing.cu:658: indexSelectLargeIndex: block: [70,0,0], thread: [0,0,0] Assertion srcIndex < srcSelectDimSize failed.

For those who encounter an error like that, I think it is because of the new version of Pytorch.
You should change 2 things.

1. In def forward(self, x).
   Add: x = x.to(torch.long)

2. in class SimpleLossCompute:
   def call(self, x, y, norm):
   Add: y = y.reshape(-1).to(torch.long)