Tal Ben-Nun comments

Results 141 comments of


                                            Tal Ben-Nun

Intellisense fails with function decorators

I can confirm that it does not happen with Pylance (i.e., the behavior is correct in pylance).

Implementation of inst2vec-imm is not available for Device Mapping task

Thanks for reporting. Yes, this will be uploaded in the next version of NCC.

loss and acc have large differences even though train and valid are set totally same

How can we reproduce this?

loss and acc have large differences even though train and valid are set totally same

Any news? Can I close this?

the links to all the datasets did not work.

@yuleihit @DapengLan thank you both for the notice! The links for the datasets expired but there is a permanent mirror at http://spclstorage.inf.ethz.ch/projects/ncc/ The links will be updated. The pretrained models...

the links to all the datasets did not work.

All the links have now been updated. Thanks again!

Expected combineable dataset

@s1530129650 That should be fine. Does anything strange happen if you comment out that line?

[train_task_classifyapp] doesn't have an opportunity to use pre-trained model weights

Sounds great. Could you upload a pull request and I'll merge?

question about predict value p at train_task_classifyapp.py line 417

Thanks for reporting this! This may be an important bug that was introduced in refactoring.

WIP: Memory pool

@targetsm after the new PR, you'll need to move the file to dace/sdfg/analysis