v1.4.0

- updated README with deprecation notice
- cap pytorch support to 1.9

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 #15
- Integrated CircleCI and tests for torch=1.7.0
- README updates for new `aten::` outputs