mindspore2022

Commit Graph

Author	SHA1	Message	Date
mindspore-ci-bot	037a121e05	!9691 expand ClipByNormNoDivSum in graph kernel From: @looop5 Reviewed-by: @gaoxiong1,@ckey_dou Signed-off-by: @ckey_dou	5 years ago
dayschan	85b69bf91f	Add a float16 restriction in the solution of reduction op's precision problem in graph splitter.	5 years ago
looop5	fa519433ef	expand ClipByNormNoDivSum	5 years ago
dayschan	297f075dca	Fix precision problem	5 years ago
zhupuxu	4f569677b7	redundant_codes Signed-off-by: zhupuxu <zhupuxu@huawei.com>	5 years ago
mindspore-ci-bot	7b311f7d2a	!9570 Modifications for GraphKernel From: @dayschan Reviewed-by: @gaoxiong1,@ckey_dou Signed-off-by: @ckey_dou	5 years ago
dayschan	6be3cc6f0d	consider atomic_add strategy in graph splitter; fixbugs; fuse and inline single op	5 years ago
looop5	848be9b07c	add tile to expand list add tile expander add BroadcastTo in model fix BroadcastTo op calling error and infer shape rewrite tile expander not split broadcast_to add SqrtGrad expander	5 years ago
dayschan	e5306b913d	GraphKernel Fuser Refactor the BasicOpsFusion and CompositeOpsFusion to one pass. Add a pass to eliminate the redundant output. TODO: rename the file basic_ops_fusion and delete the file composite_ops_fusion	5 years ago
tronzhang	2190da9946	support atomic clean and change package for akg.	5 years ago
mindspore-ci-bot	ebef1df00b	!8994 split dropout op and expand dropout From: @zengzitao Reviewed-by: Signed-off-by:	5 years ago
zengzitao	3ef0e9f053	substitute dropout by cudnnuniformreal and dropout	5 years ago
Gaoxiong	e4c3d3e0e9	update graph kernel split model	5 years ago
mindspore-ci-bot	232dff3598	!8685 [GraphKernel] For fp16 value, declare fp32 firstly and than cast to fp16 in expander From: @tronzhang Reviewed-by: Signed-off-by:	5 years ago
mindspore-ci-bot	3b946d4eb2	!8678 expand logsoftmax and grad, delete cast in softmax and fix layernorm compute dsl From: @zengzitao Reviewed-by: @gaoxiong1,@ryanww Signed-off-by: @ryanww	5 years ago
tronzhang	80f071e9fa	declare fp32 and than cast to fp16 in expander	5 years ago
tronzhang	9d7494f4df	split shape ops for more fusion pportunity.	5 years ago
zengzitao	266bfa50bf	expand logsoftmax and logsoftmax_grad, delete softmax's cast and fix layernorm op	5 years ago
zengzitao	326540cbbd	expand layernorm_grad op	5 years ago
zengzitao	28f1db74dd	expand maximum_grad minimum_grad dropout_grad op	5 years ago
dayschan	195b1fe8d5	Add Transpose into fusible list.	5 years ago
zengzitao	db27783d54	expand tanh_grad and reduce_mean, fix bug and add test_case in ci	5 years ago
zengzitao	53043ae18f	support expand fused_adam and fused_adam_weight_decay op	5 years ago
zengzitao	5cfa172720	expand gelu and gelugrad op	5 years ago
mindspore-ci-bot	5c4940cdcc	!7892 Convert non-scalar tensor to parameter Merge pull request !7892 from DeshiChen/1028_nonscalar_tensor_to_input	5 years ago
zengzitao	febdb1850c	expand bias_add and bias_add_grad op	5 years ago
dayschan	b6c2812a29	Convert non-scalar tensor to parameter Add a pass `tensor_promotion`. Fix a bug in CreateKernelInfoFromNewParameter, which reset the KernelInfo by mistake. what's more: Update akg Fixbug in model_builder when reduce axis is an interger.	5 years ago
chenzomi	44bf4c3e37	[ME] format code	5 years ago
dayschan	7599686a72	GraphKernel supports multi-output kernels	5 years ago
lingyunli63	dd48f10c3d	add assign ops in composite_topi	5 years ago
root	4e85071055	redundant codes clean	5 years ago
Gaoxiong	1cb8b803f9	update usage info	5 years ago
dayschan	37a48f6aac	GraphKernel supports GPU 1. Update akg submodule 2. Refactor akg_kernel_build, akg_ascend_kernel_build, akg_gpu_kernel_build 3. Add akg_kernel_json_decoder to support converting kernel_json to AnfNode. 4. Add GraphKernel Cost Model. (mindspore/_extends/graph_kernel) 5. Add some GraphKernel passes to GpuSession, move these passes to backend/optimizer/graph_kernel. 6. Add global id for ir files. 7. Fix bug in ConstInputToAttr.	5 years ago

33 Commits (bc84e9b4e7db505bf4c8caf508ad393f98fdfd5a)