p35189024
/
InfiniTensor
forked from jiuyuan/InfiniTensor
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity
79 Commits 70 Branches 0 Tags 141 MiB
Commit Graph

79 Commits

Author SHA1 Message Date
wendy12022 13b7a2604b
ADD add/mul/sub/div/pow operators and CPU/CUDA kernels (#26) 
Fix some

remove useless code.

add div/pow kernel

Add add/mul/sub operators.

fix cpu kernel.

add element wise kenerl for cuda.

ADD element wise operator.
 2022-09-09 13:43:59 +08:00
Anmuliar 0409eafb5f
Operators g2bmm&gbmm transplantation (#24) 
* Function tune and corresponding testcase.

*Add: Tune function in /src/kernel/cuda/conv.cc and corresponding testcase in test_conv.

*Fix: A little bug of perfRecord using in /src/core/runtime.cc.

* Tune part debug

*Add: recover the code, fixed the commit error.

*Add: some anotations in tune function

* clang formmat test

* Fix: mem leak in CUDA Runtime and Conv

* Fix: sync in conv and default sync in timeit

* Change the way to tune operator conv.

Timeit function cudNNUnfused -> Timeit function cudnnConvolutionForward.

* Change: merge the common part of cudnnunfused&tune into cudnndescriptoraccess

* clang test

* clang-format

* clang-format bash.

* Added operator G2BMM and corresponding testcase.

*Added files related to operator G2BMM creating&calling.

*Added custom_ops.cuh&custom_op.h.

* Add operator GBMML

* new version

* Fix: G2BMM and GBMM kernel bugs

* Added testcase of operator GBMML

* clang format

* Added cmake option REQUIRE_GCC9

* Delete redundent file

* Renamed class GBMML into GBMM

* clang format

* Reviewed.

* Added cudahostcompier option.

* Add: explicit CMAKE_CUDA_HOST_COMPILER

* Rename gbmm kernel

* Fix: nvcc warning in GBMM and G2BMM

Co-authored-by: wcz112 <wcz19@mails.tsinghua.edu.cn>
Co-authored-by: Liyan Zheng <liyan-zheng@outlook.com>
 2022-09-08 21:31:35 +08:00
Hardy e1d43202d7
Verify wanghailu 0902 (#22) 
* commit for verify, add some difference function

* add code for verify

* add code for verify

Co-authored-by: wanghailu <wanghailu@qiyuanlab.com>
 2022-09-05 15:45:52 +08:00
wendy12022 c3bc278c12
Op matmul (#20) 
ADD:add cuda kernel for matmul.

matmul tune

Add test_matmul.cc
 2022-09-01 21:06:55 +08:00
Hardy 32a01efbbe
add code for backtrace (#21) 
* add code for backtrace

* Add: infini::Exception

```
Test project /home/zly/InfiniTensor_aux/build
    Start 1: test_graph
1/4 Test #1: test_graph .......................   Passed    0.05 sec
    Start 2: test_hash
2/4 Test #2: test_hash ........................   Passed    0.02 sec
    Start 3: test_conv
3/4 Test #3: test_conv ........................   Passed    4.40 sec
    Start 4: test_pooling
4/4 Test #4: test_pooling .....................   Passed    2.47 sec

100% tests passed, 0 tests failed out of 4

Total Test time (real) =   6.94 sec
```

* Fix: USE_BACKTRACE in cmake

Co-authored-by: wanghailu <wanghailu@qiyuanlab.com>
Co-authored-by: Liyan Zheng <liyan-zheng@outlook.com>
 2022-09-01 20:30:12 +08:00
wendy12022 48293576c0
Add maxpool and avgpool operators (#17) 
* ADD:maxpool&&avgpool operators.

add OperatorObj::getDType()

clang format

FIX:timeit API has changed.

* Fix: Tensor::getInputs is const method

* Chore

Co-authored-by: Liyan Zheng <liyan-zheng@outlook.com>
 2022-08-31 14:44:53 +08:00
Anmuliar bd63f738dc
cuDNN conv tuning (#16) 
* Function tune and corresponding testcase.

*Add: Tune function in /src/kernel/cuda/conv.cc and corresponding testcase in test_conv.

*Fix: A little bug of perfRecord using in /src/core/runtime.cc.

* Tune part debug

*Add: recover the code, fixed the commit error.

*Add: some anotations in tune function

* clang formmat test

* Fix: mem leak in CUDA Runtime and Conv

* Fix: sync in conv and default sync in timeit

* Change the way to tune operator conv.

Timeit function cudNNUnfused -> Timeit function cudnnConvolutionForward.

* Change: merge the common part of cudnnunfused&tune into cudnndescriptoraccess

* clang test

* clang-format

* clang-format bash.

* Chore: remove print and blank lines

Co-authored-by: wcz112 <wcz19@mails.tsinghua.edu.cn>
Co-authored-by: Liyan Zheng <liyan-zheng@outlook.com>
 2022-08-29 21:37:07 +08:00
Anmuliar e076991f2f
Revert "Operator serialization (#14)" (#15) 
This reverts commit 25f0c441d2.
 2022-08-29 16:02:48 +08:00
Anmuliar 25f0c441d2
Operator serialization (#14) 
Class "Cuda Runtime" fulfills function "tune" and adds corresponding testcase.

*Add: convCudnn::tune, convCudnn::cuDNNdescriptorAccess.

*Add: testcase tune.

*Fix: a brief bug in CPU Runtime.
 2022-08-29 15:59:03 +08:00
zhengly123 93f86d3f4d
Simplify tensor transfer between CPU and CUDA (#10) 
* Add: OP infers data type  & Graph clones tensor

* Fix: vecToString format

* Add: static assert for Tensor methods

* Rename: getDataRawPtr -> getRawDataPtr

Co-authored-by: Liyan Zheng <liyan-zheng@outlook.com>
 2022-08-25 11:29:16 +08:00
zhengly123 af08df32d2
Extended DataType class and Runtime interaction (#9) 
* Add: DataType class

* Add: data-type-oblivious tensor interface

* Rename: copyBlobToCPU

Co-authored-by: Liyan Zheng <liyan-zheng@outlook.com>
 2022-08-23 16:55:59 +08:00
zhengly123 bd5934279b
Fix: rename kerels -> kernels (#8) 
Co-authored-by: Liyan Zheng <liyan-zheng@outlook.com>
 2022-08-22 16:42:55 +08:00
zhengly123 04ea5eed38
Add CUDA runtime (#6) 
* Fix: add warm-up and repetition in timing

* Add: CUDA runtime and float support

* Refactor: Cuda and Cpu runtimes inherit Runtime

* Add: environment script for Lotus

* Add: Lotus build instructions

* Update README.md

Co-authored-by: Liyan Zheng <liyan-zheng@outlook.com>
 2022-08-22 15:01:03 +08:00
zhengly123 9303ddda8e
Add Conv operator and naive CPU implemenation (#5) 
* Add: Conv definition

* Add: tensor copy data from vector

* Add: CPU conv kernel

* Fix: replace Int32 with UInt32 in DataType

Co-authored-by: Liyan Zheng <liyan-zheng@outlook.com>
 2022-08-17 14:16:01 +08:00
zhengly123 a26890abce
Tensor hash and inferShape (#4) 
* Refactor: operator hash and inferShape

* Add: hash without shape

* Add: inferShape interface for given input tensors

* Add: construct outputs in op ctor

* Add: comments for matmul

* Add: opType in AttrVector and WorkloadVector

* Chore: _graph -> graph in Op ctor

* Chore: change the "Node" suffix to "Obj"

Co-authored-by: Liyan Zheng <liyan-zheng@outlook.com>
 2022-08-15 15:08:56 +08:00
Haojie Wang eda41b06a7
Merge pull request #1 from InfiniTensor/init 
Initialization
 2022-08-09 20:21:20 +08:00
Liyan Zheng a4fb9fa413 Chore: format dbg 2022-08-09 20:16:39 +08:00
Liyan Zheng 8219b0f7ff Add: comments for Kernel 2022-08-09 20:05:01 +08:00
Liyan Zheng ce5d49c79b Add: clang format script 2022-08-09 19:50:23 +08:00
Liyan Zheng cc78a756e1 Add: clang format check github action 2022-08-09 17:58:12 +08:00
Liyan Zheng 2054b0eda4 Chore: rename getOpAttrs to getOpPerfKey 2022-08-09 15:34:28 +08:00
Liyan Zheng 8b685ae4a6 Update: OpAttrs -> OpPerfKey 2022-08-09 14:58:45 +08:00
Liyan Zheng b7e2096a26 Add: nnet code 2022-08-08 16:02:07 +08:00
Liyan Zheng 1205240218 Add: mutator abstract class 2022-08-08 15:54:17 +08:00
Liyan Zheng efa966a3e2 Add: perf engine 2022-08-07 21:12:17 +08:00
Liyan Zheng 6c356d5b42 Add: kernel registry and naive Matmul kernel 2022-08-06 15:58:40 +08:00
Liyan Zheng 559be5866d Add: Matmul operator 2022-08-05 12:50:34 +08:00
Liyan Zheng e6101b0336 Add: graph, tensor, and operator 2022-07-31 21:44:03 +08:00
Haojie Wang b89495a782
Initial commit 2022-07-27 22:40:23 +08:00