majiang31312 commented on issue #5686: URL: https://github.com/apache/incubator-tvm/issues/5686#issuecomment-635759813
I have encounted the same problem. It's triggered by 7b74a8672e1e40e7541c0007d8628586c62277e8, as vulkan use some cuda codes in topi. Forcing the sched_warp_softmax (in topi/python/topi/cuda/softmax.py) to return false could bypass the problem. The root cause of the problem seems that MakeAllreduce mistakenly assume the fourth arg should be a var node, as the var node might be replaced by a constant node in the simplify optimization. ---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org
