v0.2.4
[inference] overlap comm and compute in Linear1D_Row when stream_chunk_num > 1 (#1876)
[NFC] polish tests/test_layers/test_2d/checks_2d/check_operation_2d.py code style (#1715)
[NFC] polish test_2p5d/checks_2p5d/check_operation_2p5d.py code style (#1718)
improved allgather & reducescatter for 3d
[NFC] polish tests/test_layers/test_sequence/checks_seq/check_layer_seq.py code style (#1723)
[embedding] rename FreqAwareEmbedding -> CachedEmbedding (#1699)