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
The cuda version has been OK now and I find that the latest blocksparse from pip install version is 1.13.1, which uses cuda 10.0. But even this when I want to "from blocksparse import BlocksparseTransformer" , the module has no attribute "dw_matmul_large_n" in just the init file.
On the other hand, I 'make compile ' successfully using the commented tons of variables in the Makefile by the author. But when import, still occurs strange exceptions....
I think the init file does not mach with the public 1.13.1 blocksparse lib or build from the source? Anyone maintain this library now? Does I guess right?
I am really interested in the new paper, but this library is so hard to install or to build from source. Any update for the new blocksparse?
The text was updated successfully, but these errors were encountered: