Another reminder of the community meeting, this week! Updated agenda includes:
* Introductions
* Announcements
* TVMConf! (Chris Hoge)
* CFP closed and proposals under review
* Notifications out before end of October
* Schedule and registration open October 30
* Subproject Updates
I should emphasize that while I've used 2 convolutions to illustrate a
situation in which this technique is useful, it actually generalizes to any
operators which have some degree of locality to them (eg. max pool). In that
sense, we're not interested in matching particular well-defined patter
I think it has things to do with the complexity we want to manage in the
low-level. While it can be attractive to put a whole graph in a single TIR
block, it can inevitably increase the amount of effort to support scheduling
for such kind of blocks.
The relay represrentation is also useful to
@tqchen, So this makes me wonder -- what are the exact reasons that we need to
maintain the relay abstraction until upto the graph runtime ? As @matt-arm
mentions, I quite like the idea of making fuse-ops a TIR (the improved one with
blocks) pass because currently its forward guessing the sema
