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
We should instead print out something like slice(None, i10, None).
We plan on just having slice print out being handled inline, similar to how list/tuple. Thanks to suggestion by @t-vi & @mruberry .
Alternative
I initial thought is that we needed SliceProxy, like ListProxy / TupleProxy. I thought that's needed, but as @t-vi pointed out that
I don't think that TupleProxy, DictProxy, and ListProxy are currently not used at all, they are relicts of the functional JIT which tried to proxy everything. I think CollectionProxy is used exclusively for autograd.
Having a slice proxy is certainly possible and may be needed if we run into trouble putting numberproxies into slices.
(but I wondered why you have lists in your slices above until... ).
I don't think we need SliceProxy for now, so I'll proceed with the less invasive approach first to unblock myself.
The text was updated successfully, but these errors were encountered:
🚀 Feature
slice
is currently showing up in trace differently from list/tuple, where NumberProxy is printed explicitly, resulting in an invalid python program.e.g. in the script below:
On PR #1027, commit
4d260aed2b0939cebdeeeb4f04cf47358d3d9c8b
.We have a trace like this:
This line is not a valid python program.
We should instead print out something like
slice(None, i10, None)
.We plan on just having
slice
print out being handled inline, similar to how list/tuple. Thanks to suggestion by @t-vi & @mruberry .Alternative
I initial thought is that we needed
SliceProxy
, likeListProxy
/TupleProxy
. I thought that's needed, but as @t-vi pointed out thatI don't think we need
SliceProxy
for now, so I'll proceed with the less invasive approach first to unblock myself.The text was updated successfully, but these errors were encountered: