@areusch @junrushao1994 I have added a section to say how constants are added
to the IRModule, now.
Summary :
The storage of constants in the IRModule, will be in "Constants" attribute as
Array\
Basically, if the tir.allocate_const node is created first, then the PrimFunc
and lastly if it get
Thanks, @altanh. Your suggestion makes sense to me. To be specific, here are
two cases: parse from a python script and string.
1. When we parse from a python script, we detect the prefix `T` from the python
env (through function `__globals__`, i.e. you can even use `XXX.block` if with
`from tvm.
thanks for working through this with us @Mousius ! excited to see this land in
TVM!
--
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub:
https://github.com/apache/tvm-rfcs/pull/10#issuecomment-928239768
Merged #10 into main.
--
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub:
https://github.com/apache/tvm-rfcs/pull/10#event-5368107660
I support this change. One suggestion is that we keep a canonical namespace
that will always be parsed correctly to the TIR namespace (e.g. `tir.block`
will always work), and in particular use this identifier during printing.
Mentioning this here since I noticed the PR apache/tvm#9115 is printin
Hi @areusch ,
I have addressed candidate_memory_pool query now.
For you question around fallback :
> where are the "fallback" candidate_memory_pools passed in to the runtime?
The fallback only happens the in the compilation time as per this RFC.
Therefore, by the time USMP is done, one pool w
-1
I think we should try other measures before an enforced round-robin, such as:
- tuning the list of maintainers in a more detailed way, with perhaps less
people per directory so that lots of people don't get e-mails for simple reviews
- providing more guidance to incentive smaller, incremental
-1
It feels like the wrong solution to a valid problem. The reason I object would
be for mainly two reasons as follows :
1) The round-robin assignment could miss out interested reviewers. Giving
everyone (who have working and contributed the specific component) the
opportunity for review, IM
@areusch @tqchen ,
Could we agree to move on with using the annotations instead of AttrStmt?
--
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub:
https://github.com/apache/tvm-rfcs/pull/23#issuecomment-927950153
@areusch is there anything outstanding from you on this RFC? It seems ready to
merge after the changes you've requested.
--
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub:
https://github.com/apache/tvm-rfcs/pull/10#issuecommen
Merged #35 into main.
--
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub:
https://github.com/apache/tvm-rfcs/pull/35#event-5364638801
11 matches
Mail list logo
