Ma Zerun b017670e1b

[Improve] Use PyTorch official scaled_dot_product_attention to accelerate MultiheadAttention. (#1434) ...

* [Improve] Use PyTorch official `scaled_dot_product_attention` to accelerate `MultiheadAttention`.

* Support `--local-rank` and `--amp` option for new version PyTorch.

* Fix imports and UT.

2023-03-29 15:50:44 +08:00

..

backbones

[Improve] Use PyTorch official scaled_dot_product_attention to accelerate MultiheadAttention. (#1434)

2023-03-29 15:50:44 +08:00

classifiers

[Feature] Implement layer-wise learning rate decay optimizer constructor. (#1399)

2023-03-07 17:30:39 +08:00

heads

[Refactor] Use out_type to specify ViT-like backbone output. (#1408)

2023-03-09 11:02:58 +08:00

losses

[Refactor] Add necks, heads and losses for the self-supervised task. (#1376)

2023-02-28 10:05:00 +08:00

necks

[Refactor] Use out_type to specify ViT-like backbone output. (#1408)

2023-03-09 11:02:58 +08:00

retrievers

[Refactor] Refactor APIs, add ImageRetrievalInferencer and FeatureExtractor. (#1380)

2023-02-28 16:31:42 +08:00

selfsup

[Refactor] Use out_type to specify ViT-like backbone output. (#1408)

2023-03-09 11:02:58 +08:00

tta

…

utils

[Improve] Use PyTorch official scaled_dot_product_attention to accelerate MultiheadAttention. (#1434)

2023-03-29 15:50:44 +08:00

__init__.py

[Refactor] Add self-supervised backbones and target generators. (#1379)

2023-02-28 15:59:17 +08:00

builder.py

…