Add support for the Module class. #33

wzzju · 2021-08-19T14:03:06Z

PR types

New features

PR changes

Others

Describe

Add support for the Module class, which contains several instructions. A module can be created from a protobuf, shown as below:

ModuleProto module_proto;
// initialize the content of module_proto
// ...
Module m(module_proto);

Add the RawIterator class, which is an iterator adapter, by which we can get the raw pointer from a smart pointer. With the help of RawIterator and boost::iterator_range, we can traverse instructions in a function or functions in a module without the use of extra space. The usage demo is described as below:

for(Function &instr : module->functions()) {
// ...
}

for(Instruction &instr : function->instructions()) {
// ...
}

167: Test command: /work/Paddle/build/paddle/fluid/compiler/piano/note/note_ir_test
167: Environment variables: 
167:  FLAGS_cudnn_deterministic=true
167: Test timeout computed to be: 10000000
167: [==========] Running 6 tests from 1 test case.
167: [----------] Global test environment set-up.
167: [----------] 6 tests from IrTest
167: [ RUN      ] IrTest.ModuleToString
167: WARNING: Logging before InitGoogleLogging() is written to STDERR
167: I0823 03:43:43.420991  7334 note_ir_test.cc:221] Module string:
167: Module union_12510013719728903619
167: 
167: [[entry]] def %union_12510013719728903619(arg1.1: f32[3, 6]{}, arg2.2: f32[3, 6]{}) -> f32[3, 6]{} {
167:   %arg1.1 = f32[3, 6]{} parameter(), test_bools={true, false}, test_double=3.141000_d, test_ints={8_i, 26_i}, test_strings={"hello", "world"}
167:   %arg2.2 = f32[3, 6]{} parameter(), test_bool=true, test_doubles={5.660000_d, 6.660000_d}, test_floats={8.600000_f, 7.600000_f}, test_longs={8_l, 16_l}
167:   return %add.3 = f32[3, 6]{} add(f32[3, 6]{} %arg1.1, f32[3, 6]{} %arg2.2), test_float=-1.414000_f, test_int=-1_i, test_long=-100_l, test_string="Add"
167: }
167: 
167: [       OK ] IrTest.ModuleToString (1 ms)
167: [ RUN      ] IrTest.ModuleToProto
167: I0823 03:43:43.421424  7334 note_ir_test.cc:229] The module prototext:
167: name: "union_12510013719728903619"
167: entry_function_name: "union_12510013719728903619"
167: functions {
167:   name: "union_12510013719728903619"
167:   instructions {
167:     name: "arg1.1"
167:     opcode: "parameter"
167:     shape {
167:       element_type: F32
167:       dimensions: 3
167:       dimensions: 6
167:     }
167:     id: 1
167:     parameter_index: 0
167:     attrs {
167:       key: "test_bools"
167:       value {
167:         bools {
167:           value: true
167:           value: false
167:         }
167:       }
167:     }
167:     attrs {
167:       key: "test_double"
167:       value {
167:         d: 3.141
167:       }
167:     }
167:     attrs {
167:       key: "test_ints"
167:       value {
167:         ints {
167:           value: 8
167:           value: 26
167:         }
167:       }
167:     }
167:     attrs {
167:       key: "test_strings"
167:       value {
167:         strings {
167:           value: "hello"
167:           value: "world"
167:         }
167:       }
167:     }
167:   }
167:   instructions {
167:     name: "arg2.2"
167:     opcode: "parameter"
167:     shape {
167:       element_type: F32
167:       dimensions: 3
167:       dimensions: 6
167:     }
167:     id: 2
167:     parameter_index: 1
167:     attrs {
167:       key: "test_bool"
167:       value {
167:         b: true
167:       }
167:     }
167:     attrs {
167:       key: "test_doubles"
167:       value {
167:         doubles {
167:           value: 5.66
167:           value: 6.66
167:         }
167:       }
167:     }
167:     attrs {
167:       key: "test_floats"
167:       value {
167:         floats {
167:           value: 8.6
167:           value: 7.6
167:         }
167:       }
167:     }
167:     attrs {
167:       key: "test_longs"
167:       value {
167:         longs {
167:           value: 8
167:           value: 16
167:         }
167:       }
167:     }
167:   }
167:   instructions {
167:     name: "add.3"
167:     opcode: "add"
167:     shape {
167:       element_type: F32
167:       dimensions: 3
167:       dimensions: 6
167:     }
167:     id: 3
167:     operand_ids: 1
167:     operand_ids: 2
167:     attrs {
167:       key: "test_float"
167:       value {
167:         f: -1.414
167:       }
167:     }
167:     attrs {
167:       key: "test_int"
167:       value {
167:         i: -1
167:       }
167:     }
167:     attrs {
167:       key: "test_long"
167:       value {
167:         l: -100
167:       }
167:     }
167:     attrs {
167:       key: "test_string"
167:       value {
167:         s: "Add"
167:       }
167:     }
167:   }
167:   signature {
167:     parameters {
167:       element_type: F32
167:       dimensions: 3
167:       dimensions: 6
167:     }
167:     parameters {
167:       element_type: F32
167:       dimensions: 3
167:       dimensions: 6
167:     }
167:     parameter_names: "arg1.1"
167:     parameter_names: "arg2.2"
167:     result {
167:       element_type: F32
167:       dimensions: 3
167:       dimensions: 6
167:     }
167:   }
167:   id: 4
167:   return_id: 3
167: }
167: entry_function_signature {
167:   parameters {
167:     element_type: F32
167:     dimensions: 3
167:     dimensions: 6
167:   }
167:   parameters {
167:     element_type: F32
167:     dimensions: 3
167:     dimensions: 6
167:   }
167:   parameter_names: "arg1.1"
167:   parameter_names: "arg2.2"
167:   result {
167:     element_type: F32
167:     dimensions: 3
167:     dimensions: 6
167:   }
167: }
167: id: 4
167: entry_function_id: 4
167: [       OK ] IrTest.ModuleToProto (0 ms)
167: [ RUN      ] IrTest.ModuleDetails
167: I0823 03:43:43.422032  7334 note_ir_test.cc:249] union_12510013719728903619
167: [       OK ] IrTest.ModuleDetails (1 ms)
167: [ RUN      ] IrTest.FunctionDetails
167: I0823 03:43:43.422127  7334 note_ir_test.cc:273] arg1.1
167: I0823 03:43:43.422143  7334 note_ir_test.cc:273] arg2.2
167: I0823 03:43:43.422154  7334 note_ir_test.cc:273] add.3
167: [       OK ] IrTest.FunctionDetails (0 ms)
167: [ RUN      ] IrTest.InstructionDetails
167: [       OK ] IrTest.InstructionDetails (0 ms)
167: [ RUN      ] IrTest.VisitAttr
167: [       OK ] IrTest.VisitAttr (0 ms)
167: [----------] 6 tests from IrTest (2 ms total)
167: 
167: [----------] Global test environment tear-down
167: [==========] 6 tests from 1 test case ran. (2 ms total)
167: [  PASSED  ] 6 tests.
1/1 Test #167: note_ir_test .....................   Passed    0.01 sec

The following tests passed:
	note_ir_test

100% tests passed, 0 tests failed out of 1

class.

Add the unit test for Module.

…o note_module

thisjiang · 2021-08-23T06:21:10Z

paddle/fluid/compiler/piano/note/function.h

+  // return an instruction included in this function by the given index
+  Instruction *instruction(std::int64_t idx) const {
+    PADDLE_ENFORCE_NOT_NULL(
+        instructions_.at(idx).get(),


这儿是不是再加个判断idx < instructions_.size()比较好啊？否则at(idx)会抛异常

我们一般也是使用PADDLE_ENFORCE来检测是否下标越界，这里也正是使用at(idx)的原因。不过可以考虑要不要换成PADDLE_ENFORCE来检查。

thisjiang · 2021-08-23T06:21:12Z

paddle/fluid/compiler/piano/note/function.h

-    return param_instrs_.at(idx);
+  const Instruction &param_instr(std::int64_t idx) const {
+    PADDLE_ENFORCE_NOT_NULL(
+        return_instr_,


这儿应该是param_instrs_而不是return_instr_吧？

是的，谢谢。

thisjiang

LGTM

CtfGo · 2021-08-23T11:04:13Z

paddle/fluid/compiler/piano/note/function.h

  }

-  const Instruction *instruction(std::int64_t idx) const {
+  // return an instruction included in this function by the given index
+  Instruction *instruction(std::int64_t idx) const {


返回的instruction调用方有可能会修改吗

这是暂时不确定，我这么改之后const对象和非const对象返回的instruction在外部都是可以修改的，给了比较大的权限。你觉得需要限制吗？

需要修改啊

CtfGo · 2021-08-23T11:47:10Z

paddle/fluid/compiler/piano/note/type_traits.h

+// BaseIterator<IteratorT>::pointer
+// ...
+template <typename IteratorT>
+using BaseIterator =


BaseIterator和RawIterator名字比较泛，可以多加点限定，比如这里特指了forward的，还有IteratorT也指向的是smart ptr。另外，最好加上这个class的单测，方便后续其它开发同学大概了解怎么使用的

名字已改。这部分代码是有单测的。

int instr_cout = 0; for (Instruction& instr : func->instructions()) { instr_cout++; LOG(INFO) << instr.name(); } ASSERT_EQ(instr_cout, 3);

上述代码就是测试这个类，单独测试没法加的，毕竟其是在循环中才会使用到。

SunNy820828449 · 2021-08-23T13:02:23Z

paddle/fluid/compiler/piano/note/function.h

  }

-  const Instruction *instruction(std::int64_t idx) const {
+  // return an instruction included in this function by the given index
+  Instruction *instruction(std::int64_t idx) const {


需要修改啊

CtfGo

LGTM

* 1. add interface for fft; 2. add data type predicate; 3. fix paddle.roll. * add fft c2c cufft kernel * implement argument checking & op calling parts for fft_c2c and fftn_c2c * add operator and opmaker definitions * only register float and double for cpu. * add common code for implementing FFT, add pocketfft as a dependency * add fft c2c cufft kernel function * fix bugs in python interface * add support for c2r, r2c operators, op makers, kernels and kernel functors. * test and fix bugs * 1. fft_c2c function: add support for onesided=False; 2. add complex<float>, complex<double> support for concat and flip. * 1. fft: fix python api bugs; 2. shape_op: add support for complex data types. * fft c2c cufft kernel done with complie and link * fix shape_op, add mkl placeholder * remove mkl * complete fft c2c in gpu * 1. implement mkl-based fft, FFTC2CFunctor and common function exec_fft; 2. change the design, add input and output typename as template parameter for all FFTFunctors, update pocketfft-based implementation. * complete fft c2c on gpu in ND * complete fft c2c on gpu in ND * complete fft c2c backward in ND * fix MKL-based implementation * Add frame op and CPU/GPU kernels. * Add frame op forward unittest. * Add frame op forward unittest. * Remove axis parameter in FrameFunctor. * Add frame op grad CPU/GPU kernels and unittest. * Add frame op grad CPU/GPU kernels and unittest. * Update doc string. * Update after review and remove librosa requirement in unittest. * Update grad kernel. * add fft_c2r op * Remove data allocation in TransCompute function. * add fft r2c onesided with cpu(pocketfft/mkl) and gpu * last fft c2r functor * fix C2R and R2C for cufft, becase the direction is not an option in these cases. * add fft r2c onesided with cpu(pocketfft/mkl) and gpu * fix bugs in python APIs * fix fft_c2r grad kernal * fix bugs in python APIs * add cuda fft c2r grad kernal functor * clean code * fix fft_c2r python API * fill fft r2c result with conjugate symmetry (#19) fill fft r2c result with conjugate symmetry * add placeholder for unittests (#24) * simple parameterize test function by auto generate test case from parm list (#25) * miscellaneous fixes for python APIs (#26) * add placeholder for unittests * resize fft inputs before computation is n or s is provided. * add complex kernels for pad and pad_grad * simplify argument checking. * add type promotion * add int to float or complex promotion * fix output data type for static mode * fix fft's input dtype dispatch, import fft to paddle * fix typos in axes checking (#27) * fix typos in axes checking * fix argument checking (#28) * fix argument checking * Add C2R Python layer normal and abnormal use cases (#29) * documents and single case * test c2r case * New C2R Python layer normal and exception use cases * complete rfft,rfft2,rfftn,ihfft,ihfft2,ihfftn unittest and doc string (#30) * Documentation of the common interfaces of c2r and c2c (#31) * Documentation of the common interfaces of c2r and c2c * clean c++ code (#32) * clean code * Add numpy-based implementation of spectral ops (#33) * add numpy reference implementation of spectral ops * Add fft_c2r numpy based implementation for unittest. (#34) * add fft_c2r numpy implementation * Add deframe op and stft/istft api. (#23) * Add frame api * Add deframe op and kernels. * Add stft and istft apis. * Add deframe api. Update stft and istft apis. * Fix bug in frame_from_librosa function when input dims >= 3 * Rename deframe to overlap_add. * Update istft. * Update after code review. * Add overlap_add op and stft/istft api unittest (#35) * Add overlap_add op unittest. * Register complex kernels of squeeze/unsquuze op. * Add stft/istft api unittest. * Add unittest for fft helper functions (#36) * add unittests for fft helper functions. add complex kernel for roll op. * complete static graph unittest for all public api (#37) * Unittest of op with FFT C2C, C2R and r2c added (#38) * documents and single case * test c2r case * New C2R Python layer normal and exception use cases * Documentation of the common interfaces of c2r and c2c * Unittest of op with FFT C2C, C2R and r2c added Co-authored-by: lijiaqi <lijiaqi0612@163.com> * add fft related options to CMakeLists.txt * fix typos and clean code (#39) * fix invisible character in mkl branch and fix error in error message * clean code: remove docstring from unittest for signal.py. * always convert numpy array to paddle.Tensor to avoid comparing numpy dtype with paddle dtype. (#40) * always convert numpy array to paddle.Tensor to avoid comparing numpy dtype with paddle dtype. * fix CI Errors: numpy dtype comparison, thrust when cuda is not available (#41) 1. always convert numpy array to paddle.Tensor to avoid comparing numpy dtype with paddle dtype. 2. promote floating point tensor to complex tensor ior fft_c2c and fft_c2r; 3. fix unittest to catch UnImplementedError and RuntimeError; 4. fix compile error by avoid using thrust when cuda is not available. 5. fix sample code, use paddle.fft instead of paddle.tensor.fft * remove inclusion of thrust, add __all__ list for fft (#42) * Add api doc and update unittest. (#43) * Add doc strings. * Update overlap_add op unittest * fix MKL-based FFT implementation (#44) * fix MKL-based FFT implementation, MKL CDFT's FORWARD DOMAIN is always REAL for R2C and C2R * remove code for debug (#45) * use dynload for cufft (#46) * use std::ptrdiff_t as datatype of stride (instead of int64_t) to avoid argument mismatch on some platforms. * add complex support for fill_zeros_like * use dynload for cufft * Update doc and unittest. (#47) * Add doc of frame op and overlap_add op. * Update unittest. * use dynload for cufft (#48) 1. use dynload for cufft 2. fix unittest; 3. temporarily disable Rocm. * fix conflicts and merge upstream (#49) fix conflicts and merge upstream * fix compile error: only link dyload_cuda when cuda is available (PaddlePaddle#50) * fix compile error: only link dyload_cuda when cuda is available * fix dynload for cufft on windows (PaddlePaddle#51) 1. fix dynload for cufft on windows; 2. fix unittests. * add NOMINMAX to compile on windows (PaddlePaddle#52) add NOMINMAX to compile on windows * explicitly specify capture mode for lambdas (PaddlePaddle#55) explicitly specify capture mode for lambdas * fix fft sample (PaddlePaddle#53) * fix fft sample * update scipy and numpy version for unittests of fft (PaddlePaddle#56) update scipy and numpy version for unittests of fft * Add static graph unittests of frame and overlap_add api. (PaddlePaddle#57) * Remove cache of cuFFT & Disable ONEMKL (PaddlePaddle#59) 1. replace numpy.fft with scipy.fft as numpy<1.20 not support ortho norm 2. remove cache of cufft plans; 3. enhance error checking. 4. default WITH_ONEMKL to OFF Co-authored-by: jeff41404 <jeff41404@gmail.com> Co-authored-by: root <root@bjyz-sys-gpu-kongming9.bjyz.baidu.com> Co-authored-by: KP <109694228@qq.com> Co-authored-by: lijiaqi <lijiaqi0612@163.com> Co-authored-by: Xiaoxu Chen <chenxx_id@163.com> Co-authored-by: lijiaqi0612 <33169170+lijiaqi0612@users.noreply.github.com>

wzzju added 3 commits August 18, 2021 12:27

Update return types of some methods in Instruction class and Function

15b66fb

class.

Add methods for Module class, including ctor, ToProto, ToString, etc.

fdc8592

Add the unit test for Module.

Prohibit copying and assignment of Module class.

b243919

wzzju changed the title ~~Note module~~ Add support for Module class. Aug 20, 2021

wzzju added 4 commits August 22, 2021 05:56

Add RawIterator support for instructions() and functions().

f2dd591

Use the reference type for the range-for usage.

c3bac29

Merge branch 'paddle_compiler' of https://github.com/wzzju/Paddle int…

bab359b

…o note_module

Add more comments for Module and RawIterator.

3179d0a

wzzju marked this pull request as ready for review August 23, 2021 03:42

wzzju changed the title ~~Add support for Module class.~~ Add support for the Module class. Aug 23, 2021

thisjiang reviewed Aug 23, 2021

View reviewed changes

wzzju added 2 commits August 23, 2021 07:52

Fix some check errors and typos.

bc436dc

Add PreconditionNotMet wrappers for error messages.

46bce3a

thisjiang previously approved these changes Aug 23, 2021

View reviewed changes

CtfGo reviewed Aug 23, 2021

View reviewed changes

Rename RawIterator by UnboxingIterator.

27c335e

wzzju dismissed thisjiang’s stale review via 27c335e August 23, 2021 12:47

SunNy820828449 previously approved these changes Aug 23, 2021

View reviewed changes

Add find_if usage for func->instructions().

ccce546

wzzju dismissed SunNy820828449’s stale review via ccce546 August 23, 2021 13:15

SunNy820828449 approved these changes Aug 23, 2021

View reviewed changes

CtfGo approved these changes Aug 23, 2021

View reviewed changes

CtfGo merged commit 988e2d8 into paddle_compiler Aug 23, 2021

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Add support for the Module class. #33

Add support for the Module class. #33

wzzju commented Aug 19, 2021 •

edited

Loading

thisjiang Aug 23, 2021

wzzju Aug 23, 2021 •

edited

Loading

wzzju Aug 23, 2021

thisjiang Aug 23, 2021

wzzju Aug 23, 2021

thisjiang left a comment

CtfGo Aug 23, 2021

wzzju Aug 23, 2021

SunNy820828449 Aug 23, 2021

CtfGo Aug 23, 2021

wzzju Aug 23, 2021 •

edited

Loading

SunNy820828449 Aug 23, 2021

CtfGo left a comment

Add support for the Module class. #33

Add support for the Module class. #33

Conversation

wzzju commented Aug 19, 2021 • edited Loading

PR types

PR changes

Describe

Choose a reason for hiding this comment

wzzju Aug 23, 2021 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

thisjiang left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

wzzju Aug 23, 2021 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

CtfGo left a comment

Choose a reason for hiding this comment

wzzju commented Aug 19, 2021 •

edited

Loading

wzzju Aug 23, 2021 •

edited

Loading

wzzju Aug 23, 2021 •

edited

Loading