[GitHub] [incubator-tvm] icemelon9 commented on issue #4324: [Relay] Use opaque for where op

icemelon9 commented on issue #4324: [Relay] Use opaque for where op URL: https://github.com/apache/incubator-tvm/pull/4324#issuecomment-566804605 #4537 fix the batch_matmul schedule issue This is an automated message from the

[GitHub] [incubator-tvm] icemelon9 commented on issue #4324: [Relay] Use opaque for where op

icemelon9 commented on issue #4324: [Relay] Use opaque for where op URL: https://github.com/apache/incubator-tvm/pull/4324#issuecomment-566345060 @tqchen @kevinthesun After further investigation, I found that when the sequence length is 256, the fused batch_matmul + ones_like + where is

[GitHub] [incubator-tvm] icemelon9 commented on issue #4324: [Relay] Use opaque for where op

icemelon9 commented on issue #4324: [Relay] Use opaque for where op URL: https://github.com/apache/incubator-tvm/pull/4324#issuecomment-557844993 Agree. Close this PR for now. This is an automated message from the Apache Git

[GitHub] [incubator-tvm] icemelon9 commented on issue #4324: [Relay] Use opaque for where op

icemelon9 commented on issue #4324: [Relay] Use opaque for where op URL: https://github.com/apache/incubator-tvm/pull/4324#issuecomment-557727463 @tqchen After some benchmark, I do see there's slightly improvement on the CPU instances. On C5.9xl, with where being opaque, the latency of

[GitHub] [incubator-tvm] icemelon9 commented on issue #4324: [Relay] Use opaque for where op

icemelon9 commented on issue #4324: [Relay] Use opaque for where op URL: https://github.com/apache/incubator-tvm/pull/4324#issuecomment-556179898 Sorry about the delay. I'll look into this this week. This is an automated