Concurrently Issue All dist_feature.async_get()Requests in Base Sampler _collate_fn()#690
Concurrently Issue All dist_feature.async_get()Requests in Base Sampler _collate_fn()#690jchmura-sc wants to merge 5 commits into
dist_feature.async_get()Requests in Base Sampler _collate_fn()#690Conversation
|
/unit_test |
GiGL Automation@ 18:40:24UTC : 🔄 @ 19:45:22UTC : ✅ Workflow completed successfully. |
GiGL Automation@ 18:40:26UTC : 🔄 @ 18:50:44UTC : ✅ Workflow completed successfully. |
GiGL Automation@ 18:40:31UTC : 🔄 @ 18:42:28UTC : ✅ Workflow completed successfully. |
jchmura-sc
requested review from
kmontemayor2-sc,
mkolodner-sc,
nshah-sc,
svij-sc,
xgao4-sc,
yliu2-sc and
zfan3-sc
as code owners
kmontemayor2-sc
left a comment
kmontemayor2-sc
left a comment
There was a problem hiding this comment.
Thanks for the work Jacob! I would appreciate if we can test this at scale, is it possible to do so and follow up offline?
| futs = {} | ||
| label_keys = set() |
There was a problem hiding this comment.
nit. can we type these? e.g. futs: dict[str, torch.futures.Future[Tensor] = {}
| futs = {} | ||
| label_keys = set() |
There was a problem hiding this comment.
why do two sets of rpc send / gather here?
would it be easiest if we only sent out one batch of RPCs?
There was a problem hiding this comment.
Agreed, the heterogenous / homogenous are independent but we can simplify this!
2 participants
Close #689