p13709645
/
InfiniTensor
forked from jiuyuan/InfiniTensor
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Commit Graph

  • 40e6db6608 Add: tensor FUID in exported ONNX Liyan Zheng 2023-04-22 20:28:17 +0800
  • c451918224 Fix: tensor size overflow Liyan Zheng 2023-04-22 20:28:00 +0800
  • 34ed298725 fix format NNET_transpose whjthu 2023-04-22 17:00:52 +0800
  • 664f0dbe02 support cuda transpose whjthu 2023-04-22 16:57:27 +0800
  • 450b943dc8 add fused version in cuda/conv xxcclong 2023-04-22 16:53:03 +0800
  • a732b6f176 Fix: ignore transpose in CudaGraph since no kernel Liyan Zheng 2023-04-22 16:08:40 +0800
  • 0865f8d823 Chore: move TensorObj::clone to .cc Liyan Zheng 2023-04-22 16:03:16 +0800
  • 84f9d6731a Add: Longformer models Liyan Zheng 2023-04-22 16:00:29 +0800
  • 4f02eeb08c Add: G2BMM kernels generated by tvm 0.10 Liyan Zheng 2023-04-22 15:40:59 +0800
  • 0657938139 finish fusing conv add bias on onnx xxcclong 2023-04-22 14:40:41 +0800
  • 225a42f22d add rule for dilated conv whjthu 2023-04-21 23:40:45 +0800
  • 4e9ece76f4 Chore: remove out-of-date code Liyan Zheng 2023-04-21 23:22:40 +0800
  • 16a8c5dce5 Add: Conv1x1 rule Liyan Zheng 2023-04-21 23:21:04 +0800
  • d051460c23 Chore: suppress output Liyan Zheng 2023-04-21 22:58:18 +0800
  • d8a133684e Add: remove independent tensors in graph Liyan Zheng 2023-04-21 22:57:23 +0800
  • 9ce21200c4 Add: NMutator mode in python Liyan Zheng 2023-04-21 21:31:22 +0800
  • b943658713 Finish: GAN Liyan Zheng 2023-04-21 14:53:02 +0800
  • f021afe199 ADD unsqueeze cpu_backend2 wangw 2023-04-17 15:07:21 +0800
  • 2cd75bd79b Merge branch 'NNET_e2e_fix' into NNET_e2e Liyan Zheng 2023-04-21 13:18:44 +0800
  • f0fcbe825f Add: python verification Liyan Zheng 2023-04-21 13:07:58 +0800
  • 8c91faa948 remove expect NNET_e2e_fix huangshuhong 2023-04-21 00:17:04 +0800
  • c0ae03a2d7 fix tvm stream huangshuhong 2023-04-21 00:06:15 +0800
  • 0cb8729bc1 Add: different ONNX names for inputs and weights Liyan Zheng 2023-04-20 21:48:17 +0800
  • 8bc2d3e48d fix: test graph handler YdrMaster 2023-04-20 16:05:33 +0800
  • 28b123753e feat: 导入 Tensor 类型 YdrMaster 2023-04-20 15:26:03 +0800
  • 94730d93b5 Add: hash match for membound kernels Liyan Zheng 2023-04-20 17:16:01 +0800
  • 6d17c4caa2 Add: getPerfTime in run_models_nnet Liyan Zheng 2023-04-20 10:54:49 +0800
  • 15d0eb79cd Add: import ONNX with membound Op Liyan Zheng 2023-04-20 10:39:42 +0800
  • 2a343e240e Add: shape of intermediate tensor in exported ONNX Liyan Zheng 2023-04-20 10:28:30 +0800
  • 34ca6bf149 Fix: skip check when Graph is exported to ONNX Liyan Zheng 2023-04-19 18:10:29 +0800
  • a6019e79e3 feat(py): 支持从 Graph 直接创建 OnnxStub YdrMaster 2023-04-18 15:47:13 +0800
  • 4e1cc8d3e4 refactor(py): 使用工厂方法创建 OnnxStub YdrMaster 2023-04-18 15:32:59 +0800
  • 725f9260cf feat: 支持导出 membound YdrMaster 2023-04-19 10:33:40 +0800
  • 0edd138919 feat: 正反序列化分离为到 string 的和到 file 的 YdrMaster 2023-04-19 09:58:23 +0800
  • 0b23a065ca Add: debug hacks for InfoGAN Liyan Zheng 2023-04-19 17:23:32 +0800
  • e86e993ed4 Add: CUDA graph stream capture (MemboundOp fails) Liyan Zheng 2023-04-19 16:32:16 +0800
  • e4c20a9ae2 Add: warmup and repeat args in timeNonCtcOperators Liyan Zheng 2023-04-19 16:22:59 +0800
  • 537b3b4ea4 Add: Membound operator serialization Liyan Zheng 2023-04-18 21:53:48 +0800
  • 2812900ea2 Fix: OpType and print device tensors Liyan Zheng 2023-04-18 20:28:08 +0800
  • 01fc19795d Add: time non-compile-cime-computable operators Liyan Zheng 2023-04-18 16:08:05 +0800
  • afc4123328 Chore: remove deprecated function Liyan Zheng 2023-04-18 14:56:35 +0800
  • b981951a47 Add: NMutator::memboundToJson to export memboundOp Liyan Zheng 2023-04-18 09:56:14 +0800
  • 99b5c95455 Add: nnet::Serializer supports FuncNode Liyan Zheng 2023-04-17 20:15:40 +0800
  • 9d50b30af8 Chore: disable nnet_unimplemented_continue output Liyan Zheng 2023-04-17 13:24:56 +0800
  • bc31219bde Add: exclude compile-time computable operator time Liyan Zheng 2023-04-17 13:24:14 +0800
  • edf4e33353 Add: C++ callback to export ONNX Liyan Zheng 2023-04-17 11:48:26 +0800
  • 872f3504a9 Add: RangeOpNode::getFullExpression() Liyan Zheng 2023-04-17 11:41:13 +0800
  • da49e91ab0 Add: fuse membound operators Liyan Zheng 2023-04-17 11:38:48 +0800
  • a6b8f344d4 Chore: simplify type names Liyan Zheng 2023-04-15 13:58:18 +0800
  • 09293730ea Add: export to ONNX with custom operators Liyan Zheng 2023-04-15 00:47:11 +0800
  • 307614d95d Add: infogan python interface Liyan Zheng 2023-04-14 16:36:55 +0800
  • f14edcd52f Fix: avoid reload library Liyan Zheng 2023-04-14 15:10:47 +0800
  • d2d49c5d4f Add: invoke TVM through pipe Liyan Zheng 2023-04-13 21:43:56 +0800
  • e72fe79168 Add: search engine uses estimated time Liyan Zheng 2023-04-13 19:46:54 +0800
  • 28bf2f0e83 Add: time non-compile-cime-computable operators NNET_GAN Liyan Zheng 2023-04-18 16:08:05 +0800
  • 26f0d13c26
         Dev for 202303ddl (#66) YdrMaster 2023-04-18 15:10:33 +0800
  • 133d91bc76 Chore: remove deprecated function Liyan Zheng 2023-04-18 14:56:35 +0800
  • 37f3e9cf22 Add: NMutator::memboundToJson to export memboundOp Liyan Zheng 2023-04-18 09:56:14 +0800
  • a1974aabcd
         NNET supports TVM backend and kernels (#78) zhengly123 2023-04-18 00:26:36 +0800
  • 885a978016 Add: nnet::Serializer supports FuncNode Liyan Zheng 2023-04-17 20:15:40 +0800
  • a65dbff5f9 Chore: disable nnet_unimplemented_continue output Liyan Zheng 2023-04-17 13:24:56 +0800
  • da3f5605e5 Add: exclude compile-time computable operator time Liyan Zheng 2023-04-17 13:24:14 +0800
  • 43d4798323
         ADD: sub graph replacement. (#56) test_onnx wendy12022 2023-04-17 13:09:07 +0800
  • c8b2c8ed32
         Cpu backend2 (#77) wendy12022 2023-04-17 12:15:23 +0800
  • 45f7841d83 Add: C++ callback to export ONNX Liyan Zheng 2023-04-17 11:48:26 +0800
  • 7a1d271c79 Add: RangeOpNode::getFullExpression() Liyan Zheng 2023-04-17 11:41:13 +0800
  • b2c53458d9 Add: fuse membound operators Liyan Zheng 2023-04-17 11:38:48 +0800
  • e76f2d0f10 Chore: simplify type names Liyan Zheng 2023-04-15 13:58:18 +0800
  • f881db7ceb Add: export to ONNX with custom operators Liyan Zheng 2023-04-15 00:47:11 +0800
  • b72662bb9b Add: infogan python interface Liyan Zheng 2023-04-14 16:36:55 +0800
  • 582de83629 Fix: avoid reload library Liyan Zheng 2023-04-14 15:10:47 +0800
  • b6b37ccf33 Add: invoke TVM through pipe Liyan Zheng 2023-04-13 21:43:56 +0800
  • 9a009be337 Add: search engine uses estimated time Liyan Zheng 2023-04-13 19:46:54 +0800
  • f0785225d7 Chore: format and comments Liyan Zheng 2023-04-12 11:45:59 +0800
  • f939ebf8bb Add: comments for Matmul Liyan Zheng 2023-04-12 11:19:09 +0800
  • 31b03ef91a Add: detailed output of Matmul Liyan Zheng 2023-04-11 16:51:42 +0800
  • bd76a5c8d6 Add: CMake support of TVM Liyan Zheng 2023-04-11 16:42:50 +0800
  • 83d5842f56 Fix: specify runtime Liyan Zheng 2023-04-11 16:33:54 +0800
  • 005c113948 Add: support TVM packed function Liyan Zheng 2023-04-11 15:49:43 +0800
  • 0d70643867 Add: Random generator Liyan Zheng 2023-04-11 15:36:35 +0800
  • f29f54687c Add: PRelu Expr and AsTVMVisitor Liyan Zheng 2023-04-10 23:30:34 +0800
  • 67e9fb9632 Fix: test_memboundOp huangshuhong 2023-04-11 23:40:41 +0800
  • af80830d56 Fix: membound_tvm huangshuhong 2023-04-10 22:21:48 +0800
  • f19bd4d1d6 Fix: remove cuCtxCreate which makes TVM fails huangshuhong 2023-04-10 15:11:34 +0800
  • 2e501de8db Add: PackedFunc (broken) huangshuhong 2023-04-10 13:26:33 +0800
  • 31b1edc600 Add: CMake config Liyan Zheng 2023-04-09 17:17:10 +0800
  • 69d894e003 Add: TVM headers and CMake include paths Liyan Zheng 2023-04-09 17:16:36 +0800
  • e8b4e3f03f Add: expr simplication for TVM Liyan Zheng 2023-04-07 13:48:29 +0800
  • 0d8685b1e0 Fix: graph OpList ctor Liyan Zheng 2023-04-06 17:14:00 +0800
  • 04858012bf Add: Exception for checkCuError Liyan Zheng 2023-04-06 17:12:41 +0800
  • 6aaf963f50 Fix: cuBLAS failure when codegen is enabled Liyan Zheng 2023-03-29 12:52:04 +0800
  • 6441abb88f Fix: GraphObj ctor should use cloneTensor Liyan Zheng 2023-03-14 21:32:15 +0800
  • 220b76f355 Add: matmul broadcast huangshuhong 2023-03-14 20:17:16 +0800
  • aac4bd4650 Fix: [Intermediate] NMutator::expressionToGraph Liyan Zheng 2023-03-12 21:51:39 +0800
  • 7478b698dc Add: expression reader as a cmake target Liyan Zheng 2023-03-06 20:33:39 +0800
  • 8e6fc4136e Add: cache and padding (bugs!!) huangshuhong 2023-03-05 22:50:31 +0800
  • bd12b67ea3 Add: mutator InfoGAN minimum test Liyan Zheng 2023-03-03 20:58:29 +0800
  • f2be122efc change Device::MKL to Device::INTELCPU wangw 2022-11-25 17:55:23 +0800
  • fe1afe38fa
         fix code of bang conv (#76) Hardy 2023-03-29 15:47:32 +0800
  • 823e66a9ff
         Support perf bang 1115 (#57) Hardy 2023-03-29 13:52:56 +0800