v0.2.4
[hotfix] fix a running error in test_colo_checkpoint.py (#1387)
[CheckpointIO] a uniform checkpoint I/O module (#1689)
Refactored docstring to google style
[gemini] fix colo_init_context (#2683)
[setup] support pre-build and jit-build of cuda kernels (#2374)
[Gemini] clean no used MemTraceOp (#1970)
[pipeline/rank_recorder] fix bug when process data before backward | add a tool for multiple ranks debug (#1681)
[NFC] polish utils/tensor_detector/__init__.py code style (#1573)
[ddp] add is_ddp_ignored (#2434)
[utils] Add use_reetrant=False in utils.activation_checkpoint (#1460)
[utils] refactor parallel layers checkpoint and bcast model on loading checkpoint (#1548)
[refactor] refactor the memory utils (#715)
[gemini] APIs to set cpu memory capacity (#809)