Kulin Seth
Kulin Seth
> @kulinseth @albanD I noticed this ticket's been subject to triage, and a few other folks have filed issues regarding similar observations. > > Do you know how the [figure...
> @kulinseth thanks for sharing, I'll try running that on my system, and I'm curious to dig in, because when I ran my own benchmarks with HuggingFace's BERT on an...
> Device: M1 Pro 10CPU 16GPU > > Training my model using mps: > > ``` > 0m 2s (Time Remaining: 4167m 10s) (epoch: 1 finish: 0%) > 0m 5s...
@demosn01 currently the graphs for random ops are not cached, which is leading to un-necessary recompiles slowing the randn operations. We are actively working on it and will update here...
> +1 on Alban's feedback: should this change include some testing? Agreed, the testConsistency `nn.functional.linear` still has some issues. I will enable a subset which was fixed. Done. Enabled the...
> +1 on Alban's feedback: should this change include some testing? @malfet , can you please take a look now.
> Added `ciflow/mps` tag to schedule just the tests we care about (plan to add action that auto assigns this label soon) Sound great.
@pytorchbot merge
@msaroufim , are there any plans to add the M1 workflows for PyTorch examples ? We have some machines used in ```ciflow/mps``` for pytorch/pytorch project. cc @malfet
@hudeven , is there a bot merge command to check this in ?