Balint Cristian

Daniel, - Could please tell me where you add this modification ? I am also interested in this. We could add a special config parameter like "tied_transpose_to:" or something like.

I've manged it. Will come back a bit later with a patch proposal, and with results over "ae" example. Will post it over issue #48.

- It is nonsense to tie a transposed bias/weight (e.g in case of autoencoder) to some other layer in the very learning process. However I manage it in the code...

Hi @lhutton1 ! Thanks a lot for picking up the recently introduce llvm reflection (well, kind of) for ARM targets too ! > Currently, target features are determined by a...

> It seems CI was failing due to a memory leak observed when calling `GetAllLLVMCpuFeatures()` and `GetAllLLVMTargetArches()`. The following is a valgrind report for an executable that creates a new...

> Thanks for the quick response @cbalint13! I didn't try with llvm18 yet, only llvm17. Calling GetAllLLVMCpuFeatures() and GetAllLLVMTargetArches() should reproduce it, but I'll be able to come up with...

> Here is a reproducer: > mem_leak.cpp Thanks a lot for this, I start to look at it now.

> > Here is a reproducer: > > mem_leak.cpp > > Thanks a lot for this, I start to look at it now. @lhutton1 , Here is a patch: [tvm-llvm-memleak.diff.gz](https://github.com/apache/tvm/files/14762596/tvm-llvm-memleak.diff.gz)...

> > > Here is a reproducer: > > > mem_leak.cpp > > > > > > Thanks a lot for this, I start to look at it now. >...

> This PR causes spurious error messages to be printed when loading `libtvm.so`, when the local version of LLVM available does not support one of the CPU architectures defined in...