So long and thanks for all the fish

Please refer to the official profiler that has been redesigned in PyTorch 1.9.
https://pytorch.org/docs/1.9.0/profiler.html?highlight=profiler#module-torch.profiler

v1.3.4

v1.3.4

v1.3.3

v1.3.3

v1.3.2

- Addressed swapped self memory usage columns in table output

v1.3.1

- non sequential events are no longer dropped, `group_by` is still O(n) (see issue #17)

v1.3.0

- Added in support for `profile_memory` (torch >= 1.6), with runtime warning if
unsupported
- Added in support for `self_cuda_total` (torch >= 1.7), graceful
fallback if unsupported
- refactored display and profile code into separate files
- dynamically populating table columns based on options
- table column headings consistent with official names

v1.2.0

- Added in @CorentinJ PR https://github.com/awwong1/torchprof/pull/15
- Integrated CircleCI and tests for torch=1.7.0
- README updates for new `aten::` outputs

v1.1.1

v1.1.1

v1.1.0

v1.1.0

v1.0.0

v1.0.0