David Berard

Results 27 issues of David Berard

Stack from [ghstack](https://github.com/ezyang/ghstack): * __->__ #83130 Copied from @jamesr66a 's example in #83116. Implements `__deepcopy__` to skip deepcopying the elements of `_autowrap_search`, because it contains modules, which cannot/should not be...

cla signed
fx

Stack from [ghstack](https://github.com/ezyang/ghstack): * __->__ #870 * #801 Currently failing NNC https://github.com/pytorch/pytorch/issues/75925. Once it is fixed in NNC, we can re-enable it in the microbenchmarks suite

cla signed

Previously, we would launch a different slurm job for measurement, e.g.: 1) resnet50 w/ inductor + graph breaks, 2 node 2) resnet50 w/ inductor + NO graph breaks, 2 node...

cla signed

Stack from [ghstack](https://github.com/ezyang/ghstack): * __->__ #1243 Many of the models are fairly large, and 10 iterations is way too much (chrometrace take sa long time to load and freezes). 2...

cla signed

Can model it after https://github.com/pytorch/torchdynamo/blob/f11a33107407c51e93cba96e26c2f65e56edbc1d/torchdynamo/testing.py#L37

Stack from [ghstack](https://github.com/ezyang/ghstack) (oldest at bottom): * __->__ #1173 Copied from this branch: https://github.com/pytorch/benchmark/tree/wconstab/ddp_experiments This is necessary to make the optimizer capturable by dynamo.

cla signed

Stack from [ghstack](https://github.com/ezyang/ghstack): * __->__ #93043 for some tensor x, x.type(torch.FloatTensor) will essentially do the same thing as x.to(torch.float). x.type can be called with at least 3 types of inputs:...

topic: bug fixes
module: dynamo
ciflow/inductor
release notes: dynamo

Stack from [ghstack](https://github.com/ezyang/ghstack) (oldest at bottom): * __->__ #120141

ciflow/inductor

Stack from [ghstack](https://github.com/ezyang/ghstack) (oldest at bottom): * __->__ #119707 **Motivation**: Previously, we had an issue where the number of forward outputs wouldn't match the metadata in ViewAndMutationMetadata. (This would cause...

module: dynamo
ciflow/inductor
release notes: AO frontend

Stack from [ghstack](https://github.com/ezyang/ghstack) (oldest at bottom): * __->__ #120136 cc @voznesenskym @penguinwu @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @wenzhe-nrv @jiayisunx @chenyang78 @aakhundov @kadeng

module: dynamo
ciflow/inductor
release notes: AO frontend