pengwa 5eda79bdd3 Improve perf for stage3 training (#18099) ... ### Improve perf for stage3 training - first wave Port existing PythonOp/PythonOpGrad python runner to C++, also introduce an unsafe run mode (to skip inplace, save for backward, materrialized grad detection on the fly). This reduce the overhead from XX~XXX us to X ~ lower end of XX us . In LLAMA2 7B training with 8x32GV100, we have observed 6.7% gains over PyTorch. (1.59 v.s. 1.49it/s) Peak memory also dropped from 31GB to 28GB. ### Motivation and Context <!-- - Why is this change required? What problem does it solve? - If it fixes an open issue, please link to the issue here. -->		2023-12-15 13:32:19 +08:00
..
core	Improve perf for stage3 training (#18099)	2023-12-15 13:32:19 +08:00
lazy_tensor	NVTX windows include and link fixes (#16831)	2023-08-16 11:53:58 -07:00
models	Handle string initializers in constant folding (#17422)	2023-09-27 21:15:58 +10:00
python	Improve perf for stage3 training (#18099)	2023-12-15 13:32:19 +08:00
test	Improve perf for stage3 training (#18099)	2023-12-15 13:32:19 +08:00
training_api	[On-Device Training] Expose Parameters through the Training API (#17364)	2023-09-25 20:03:24 -07:00
training_ops	Improve perf for stage3 training (#18099)	2023-12-15 13:32:19 +08:00