Tianlei Wu 5dce9be4f9 Add Attention Fusion Transformer (#2445) ... Add Attention Fusion Transformer to fuse multi-head self attention subgraph to one node for optimizing Bert model inference performance. It supports BERT model exported from PyTorch. It fuses about 20 nodes into one Attention node, and could significantly improve the inference speed of BERT model. Support symbolic (first dimension for batch size) in input shape.		2019-11-25 14:46:37 -08:00
..
automl_ops
common	Move CXX API global into the header (#2228)	2019-10-23 14:15:53 -07:00
contrib_ops	Fuse SkipLayerNorm with Bias (#2453)	2019-11-22 14:20:46 -08:00
framework	Clean up build.py (#2446)	2019-11-22 12:14:03 -08:00
ir	Avoid using the default logger in the graph lib and optimizers (#2361)	2019-11-14 13:23:28 -08:00
mlas	MLAS: Implement U8S8 GEMV kernels (#2069)	2019-10-09 11:54:16 -07:00
onnx	Clean up build.py (#2446)	2019-11-22 12:14:03 -08:00
opaque_api	Avoid using the default logger in the graph lib and optimizers (#2361)	2019-11-14 13:23:28 -08:00
optimizer	Add Attention Fusion Transformer (#2445)	2019-11-25 14:46:37 -08:00
perftest	[v2] Add ACL (Arm Compute Library) execution provider (#2258)	2019-10-31 12:25:36 -07:00
platform	Fixed tensor reference to const data and cleaned up Env API. (#1979)	2019-10-24 10:28:13 -07:00
proto
providers	Correctly handle implicit inputs for fused nodes (#2390)	2019-11-21 10:27:09 -08:00
python	Symbolic shape inference improvements: (#2460)	2019-11-25 10:31:13 -08:00
server	Update ORTSRV integration test model path (#2225)	2019-10-23 13:48:20 -07:00
shared_lib	CSharp api and test for loading custom op shared library (#2420)	2019-11-21 15:45:49 -08:00
testdata	Add Attention Fusion Transformer (#2445)	2019-11-25 14:46:37 -08:00
tvm
util	Avoid using the default logger in the graph lib and optimizers (#2361)	2019-11-14 13:23:28 -08:00
win_getopt