imperative.cc 118.1 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
/* Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.

Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at

http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License. */

#include "paddle/fluid/pybind/imperative.h"
16

17
#include <Python.h>
18 19 20 21
#include <pybind11/chrono.h>
#include <pybind11/complex.h>
#include <pybind11/functional.h>
#include <pybind11/stl.h>
22

23
#include <algorithm>
24
#include <memory>
25
#include <set>
J
Jiabin Yang 已提交
26
#include <string>
27
#include <unordered_map>
28
#include <unordered_set>
29
#include <utility>
J
Jiabin Yang 已提交
30
#include <vector>
31

32
#include "paddle/fluid/framework/scope_guard.h"
33
#include "paddle/fluid/imperative/all_reduce.h"
34
#include "paddle/fluid/imperative/amp_auto_cast.h"
35
#include "paddle/fluid/imperative/basic_engine.h"
36
#include "paddle/fluid/imperative/bkcl_context.h"
37
#include "paddle/fluid/imperative/data_loader.h"
38
#include "paddle/fluid/imperative/gloo_context.h"
39
#include "paddle/fluid/imperative/hooks.h"
40
#include "paddle/fluid/imperative/layer.h"
J
Jiabin Yang 已提交
41
#include "paddle/fluid/imperative/nccl_context.h"
42
#include "paddle/fluid/imperative/partial_grad_engine.h"
43
#include "paddle/fluid/imperative/profiler.h"
44
#include "paddle/fluid/imperative/py_layer_fwd.h"
45
#include "paddle/fluid/imperative/reducer.h"
46
#include "paddle/fluid/imperative/tracer.h"
M
minqiyang 已提交
47
#include "paddle/fluid/imperative/type_defs.h"
48
#include "paddle/fluid/memory/allocation/mmap_allocator.h"
49
#include "paddle/fluid/operators/utils.h"
50
#include "paddle/fluid/pybind/op_function.h"
51
#include "paddle/fluid/pybind/pybind_boost_headers.h"
L
Leo Chen 已提交
52
#include "paddle/fluid/pybind/tensor_py.h"
53

54 55 56
namespace paddle {
namespace pybind {

57 58
PyTypeObject *g_varbase_pytype = nullptr;

59 60
namespace py = ::pybind11;

61 62 63 64
class Layer : public imperative::Layer {
 public:
  using imperative::Layer::Layer;  // Inherit constructors

65 66 67 68
  std::vector<std::shared_ptr<imperative::VarBase>> Forward(
      const std::vector<std::shared_ptr<imperative::VarBase>> &inputs)
      override {
    PYBIND11_OVERLOAD(std::vector<std::shared_ptr<imperative::VarBase>>, Layer,
J
Jiabin Yang 已提交
69
                      Forward, inputs);  // NOLINT
70 71 72
  }
};

73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131
template <typename T>
static T PyObjectCast(PyObject *obj) {
  try {
    return py::cast<T>(py::handle(obj));
  } catch (py::cast_error &) {
    PADDLE_THROW(platform::errors::InvalidArgument(
        "Python object is not type of %s", typeid(T).name()));
  }
}

class PyVariableWrapperHook : public imperative::VariableWrapperHook {
 public:
  explicit PyVariableWrapperHook(PyObject *func) : py_func_(func) {
    Py_INCREF(py_func_);
  }

  ~PyVariableWrapperHook() {
    py::gil_scoped_acquire gil;
    Py_DECREF(py_func_);
  }

  std::shared_ptr<imperative::VariableWrapper> operator()(
      const std::shared_ptr<imperative::VariableWrapper> &var) override {
    py::gil_scoped_acquire gil;
    VLOG(3) << "Call PyVariableWrapperHook for var " << var->Name();

    // 1. unpack temp VarBase from VariableWrapper
    std::shared_ptr<imperative::VarBase> tmp_varbase =
        std::make_shared<imperative::VarBase>(var);

    // 2. call hook and return
    PyObject *res = nullptr;
    try {
      res = PyObject_CallFunctionObjArgs(py_func_, py::cast(tmp_varbase).ptr(),
                                         nullptr);
    } catch (platform::EnforceNotMet &e) {
      throw std::move(e);
    } catch (std::exception &e) {
      PADDLE_THROW(platform::errors::Unavailable(
          "Hook function of Tensor raises an exception: %s.", e.what()));
    } catch (...) {
      PADDLE_THROW(platform::errors::Fatal(
          "Hook function of Tensor raises an unknown exception."));
    }

    PADDLE_ENFORCE_NOT_NULL(res,
                            platform::errors::Unavailable(
                                "Hook function of Tensor return a nullptr."));
    if (res == Py_None) {
      return var;
    }

    return PyObjectCast<std::shared_ptr<imperative::VarBase>>(res)->SharedVar();
  }

 private:
  PyObject *py_func_;
};

L
Leo Chen 已提交
132 133 134 135 136
static const platform::Place PyObjectToPlace(const py::object &place_obj) {
  if (py::isinstance<platform::CPUPlace>(place_obj)) {
    return place_obj.cast<platform::CPUPlace>();
  } else if (py::isinstance<platform::CUDAPlace>(place_obj)) {
    return place_obj.cast<platform::CUDAPlace>();
137 138
  } else if (py::isinstance<platform::XPUPlace>(place_obj)) {
    return place_obj.cast<platform::XPUPlace>();
L
Leo Chen 已提交
139 140
  } else if (py::isinstance<platform::CUDAPinnedPlace>(place_obj)) {
    return place_obj.cast<platform::CUDAPinnedPlace>();
141 142
  } else if (py::isinstance<platform::NPUPlace>(place_obj)) {
    return place_obj.cast<platform::NPUPlace>();
143 144
  } else if (py::isinstance<platform::Place>(place_obj)) {
    return place_obj.cast<platform::Place>();
L
Leo Chen 已提交
145 146
  } else {
    PADDLE_THROW(platform::errors::InvalidArgument(
147
        "Place should be one of "
148
        "Place/CPUPlace/XPUPlace/CUDAPlace/CUDAPinnedPlace/NPUPlace"));
L
Leo Chen 已提交
149 150 151
  }
}

L
Leo Chen 已提交
152 153 154 155 156 157 158 159 160 161
// only initialize varbase, but not its tensor.
static void InitVarBaseOnly(imperative::VarBase *self, const std::string &name,
                            bool persistable = false, int stop_gradient = -1) {
  auto name_ = name == ""
                   ? imperative::GetCurrentTracer()->GenerateUniqueName(
                         "generated_tensor")
                   : name;

  VLOG(5) << "Init Tensor as: / name: " << name_
          << " / persistable: " << persistable
162
          << " / stop_gradient: " << stop_gradient;
L
Leo Chen 已提交
163 164 165 166 167 168 169 170 171 172 173 174 175 176
  new (self) imperative::VarBase(name_);
  if (stop_gradient != -1) {
    self->SetOverridedStopGradient(stop_gradient);
  }
  self->SetPersistable(persistable);
  self->SetType(framework::proto::VarType::LOD_TENSOR);
}

// initialize varbase and its tensor.
static void InitVarBaseAndTensor(
    imperative::VarBase *self, const py::array &array,
    const platform::Place &place, const std::string &name,
    bool persistable = false, bool zero_copy = false, int stop_gradient = -1) {
  InitVarBaseOnly(self, name, persistable, stop_gradient);
177
  auto *tensor = self->MutableVar()->GetMutable<framework::LoDTensor>();
L
Leo Chen 已提交
178
  VLOG(4) << "zero_copy: " << zero_copy;
L
Leo Chen 已提交
179 180
  if (platform::is_cpu_place(place)) {
    SetTensorFromPyArray<platform::CPUPlace>(
181
        tensor, array, BOOST_GET_CONST(platform::CPUPlace, place), zero_copy);
182 183 184
  } else if (platform::is_xpu_place(place)) {
    SetTensorFromPyArray<platform::XPUPlace>(
        tensor, array, BOOST_GET_CONST(platform::XPUPlace, place), zero_copy);
L
Leo Chen 已提交
185 186
  } else if (platform::is_gpu_place(place)) {
    SetTensorFromPyArray<platform::CUDAPlace>(
187
        tensor, array, BOOST_GET_CONST(platform::CUDAPlace, place), zero_copy);
L
Leo Chen 已提交
188 189
  } else if (platform::is_cuda_pinned_place(place)) {
    SetTensorFromPyArray<platform::CUDAPinnedPlace>(
190 191
        tensor, array, BOOST_GET_CONST(platform::CUDAPinnedPlace, place),
        zero_copy);
192 193 194
  } else if (platform::is_npu_place(place)) {
    SetTensorFromPyArray<platform::NPUPlace>(
        tensor, array, BOOST_GET_CONST(platform::NPUPlace, place), zero_copy);
195
  } else {
L
Leo Chen 已提交
196
    PADDLE_THROW(platform::errors::InvalidArgument(
197 198
        "Place should be one of "
        "CPUPlace/XPUPlace/CUDAPlace/CUDAPinnedPlace/NPUPlace"));
J
Jiabin Yang 已提交
199
  }
200 201 202 203 204
  self->SetDataType(tensor->type());
}

static void InitVarBaseFromNumpyWithKwargs(imperative::VarBase *self,
                                           const py::kwargs &kwargs) {
205
  VLOG(4) << "Init VarBase from kwargs: ";
L
Leo Chen 已提交
206 207 208 209 210 211
  auto persistable = kwargs.contains("persistable")
                         ? kwargs["persistable"].cast<bool>()
                         : false;
  auto zero_copy =
      kwargs.contains("zero_copy") ? kwargs["zero_copy"].cast<bool>() : false;
  auto name = kwargs.contains("name") ? kwargs["name"].cast<std::string>() : "";
212 213 214
  auto stop_gradient = kwargs.contains("stop_gradient")
                           ? kwargs["stop_gradient"].cast<int>()
                           : -1;
L
Leo Chen 已提交
215
  auto default_place = imperative::GetCurrentTracer()->ExpectedPlace();
L
Leo Chen 已提交
216 217 218 219 220 221 222 223 224 225 226 227

  if (kwargs.contains("value")) {
    auto array = kwargs["value"].cast<py::array>();
    // place is only used when array is given, otherwise, it is meaningless and
    // ignored
    auto place = kwargs.contains("place") ? PyObjectToPlace(kwargs["place"])
                                          : default_place;
    InitVarBaseAndTensor(self, array, place, name, persistable, zero_copy,
                         stop_gradient);
  } else {
    InitVarBaseOnly(self, name, persistable, stop_gradient);
  }
228
}
229

230 231 232
template <typename P>
static void InitVarBaseFromNumpyWithArg(imperative::VarBase *self,
                                        const py::array &array, const P &place,
L
Leo Chen 已提交
233 234
                                        bool persistable = false,
                                        bool zero_copy = false,
235 236 237 238 239
                                        std::string name = "",
                                        int stop_gradient = -1) {
  VLOG(4) << "Init VarBase from Arg: ";
  // 0: self, 1: value, 2: place, 3: persistable, 4: zero_copy, 5: name , 6:
  // stop_gradient
L
Leo Chen 已提交
240
  if (name == "") {
241 242
    name =
        imperative::GetCurrentTracer()->GenerateUniqueName("generated_tensor");
L
Leo Chen 已提交
243
  }
244 245
  VLOG(5) << "Init Tensor as: / name: " << name
          << " / persistable: " << persistable << " / zero_copy: " << zero_copy
246
          << " / stop_gradient: " << stop_gradient << " / at " << place;
L
Leo Chen 已提交
247
  new (self) imperative::VarBase(name);
248 249
  self->SetPersistable(persistable);
  auto *tensor = self->MutableVar()->GetMutable<framework::LoDTensor>();
250 251 252
  if (stop_gradient != -1) {
    self->SetOverridedStopGradient(stop_gradient);
  }
253 254 255 256 257 258
  SetTensorFromPyArray<P>(tensor, array, place, zero_copy);
  self->SetType(framework::proto::VarType::LOD_TENSOR);
  self->SetDataType(tensor->type());
}

static void InitVarBaseFromNumpyWithArgDefault(imperative::VarBase *self,
L
Leo Chen 已提交
259 260
                                               const py::array &array) {
  auto place = imperative::GetCurrentTracer()->ExpectedPlace();
261
  VLOG(4) << "Init VarBase from numpy at " << place;
L
Leo Chen 已提交
262
  InitVarBaseAndTensor(self, array, place, "");
263
}
264

265
static void InitVarBaseFromTensorWithArgDefault(
266
    imperative::VarBase *self, const framework::Tensor &tensor) {
267 268 269
  VLOG(4) << "Init VarBase";
  auto place = imperative::GetCurrentTracer()->ExpectedPlace();
  new (self) imperative::VarBase(
270
      imperative::GetCurrentTracer()->GenerateUniqueName("generated_tensor"));
271 272 273 274 275 276 277 278 279 280 281 282 283 284
  self->SetPersistable(false);
  self->SetType(framework::proto::VarType::LOD_TENSOR);
  self->SetDataType(tensor.type());
  auto *new_tensor = self->MutableVar()->GetMutable<framework::LoDTensor>();
  // Same place,share data directly
  if (place == tensor.place()) {
    new_tensor->ShareDataWith(tensor);
    VLOG(4) << "Same place, do ShareDataWith";
  } else {
    framework::TensorCopy(tensor, place, new_tensor);
    VLOG(4) << "Different place, do TensorCopy";
  }
}

285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305
template <typename P>
static void InitVarBaseFromTensorWithArg(imperative::VarBase *self,
                                         const framework::Tensor &tensor,
                                         const P &place) {
  VLOG(4) << "Init VarBase";
  new (self) imperative::VarBase(
      imperative::GetCurrentTracer()->GenerateUniqueName("generated_tensor"));
  self->SetPersistable(false);
  self->SetType(framework::proto::VarType::LOD_TENSOR);
  self->SetDataType(tensor.type());
  auto *new_tensor = self->MutableVar()->GetMutable<framework::LoDTensor>();
  // Same place,share data directly
  if (platform::is_same_place(place, tensor.place())) {
    new_tensor->ShareDataWith(tensor);
    VLOG(4) << "Same place, do ShareDataWith";
  } else {
    framework::TensorCopy(tensor, place, new_tensor);
    VLOG(4) << "Different place, do TensorCopy";
  }
}

306 307 308 309 310
static std::string GetTypeName(const imperative::VarBase &var) {
  if (var.Type() == framework::proto::VarType::RAW) {
    return "RAW";
  } else if (!var.Var().IsInitialized()) {
    return "nullptr";
311
  } else {
312
    return framework::ToTypeName(var.Var().Type());
313 314
  }
}
L
Leo Chen 已提交
315

316
using PyNameVarBaseMap = std::unordered_map<std::string, py::handle>;
317 318 319 320 321 322 323 324 325 326 327 328 329

// NOTE(zjl): py::handle is a very light wrapper of PyObject *.
// Unlike py::object, py::handle does not change reference count of PyObject *.
static std::vector<std::shared_ptr<imperative::VarBase>>
GetVarBaseListFromPyHandle(const py::handle &handle) {
  PyObject *py_obj = handle.ptr();  // get underlying PyObject
  // Python None is not nullptr in C++!
  if (!py_obj || py_obj == Py_None) {
    return {};
  }

  std::vector<std::shared_ptr<imperative::VarBase>> result;

330
  if (PyList_Check(py_obj)) {  // List of VarBase
331 332 333
    size_t len = PyList_GET_SIZE(py_obj);
    result.reserve(len);
    for (size_t i = 0; i < len; ++i) {
334 335 336
      PyObject *py_ivar = PyList_GET_ITEM(py_obj, i);
      PADDLE_ENFORCE_NOT_NULL(
          py_ivar, platform::errors::InvalidArgument("Python Object is NULL"));
337 338 339
      result.emplace_back(
          PyObjectCast<std::shared_ptr<imperative::VarBase>>(py_ivar));
    }
340
  } else if (PyTuple_Check(py_obj)) {  // Tuple of VarBase
341 342 343
    size_t len = PyTuple_GET_SIZE(py_obj);
    result.reserve(len);
    for (size_t i = 0; i < len; ++i) {
344 345 346
      PyObject *py_ivar = PyTuple_GET_ITEM(py_obj, i);
      PADDLE_ENFORCE_NOT_NULL(
          py_ivar, platform::errors::InvalidArgument("Python Object is NULL"));
347 348 349
      result.emplace_back(
          PyObjectCast<std::shared_ptr<imperative::VarBase>>(py_ivar));
    }
350 351 352
  } else {  // VarBase
    result.emplace_back(
        PyObjectCast<std::shared_ptr<imperative::VarBase>>(py_obj));
353 354 355 356
  }

  return result;
}
357 358 359 360 361 362 363 364
static bool IsNumpyType(PyObject *obj) {
  // It is not a good way to judge the type of obj by its type'name. Maybe using
  // `PyArray_IsScalar` will be better. However, this interface cannot be used
  // by including pybind11, and it needs to compile with numpy.
  auto type_name = std::string(Py_TYPE(obj)->tp_name);
  return type_name == "numpy.int64" || type_name == "numpy.longlong" ||
         type_name == "numpy.int32" || type_name == "numpy.int16";
}
365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409

static bool PyCheckTensor(PyObject *obj) {
  return py::isinstance<imperative::VarBase>(obj);
}

// cast numpy type form S to T, this may allocate new memory
template <class T, class S>
static py::array_t<T> CastNumpyType(py::array_t<S> array) {
  if (std::is_same<T, S>::value) {
    return array;
  }
  auto dim = array.ndim();
  std::vector<py::ssize_t> result_shape(dim);
  for (auto i = 0; i < dim; i++) {
    result_shape[i] = array.shape(i);
  }

  py::array_t<T> result(result_shape);

  return py::vectorize([](S s) { return static_cast<T>(s); })(array);
}

template <class T>
static py::array_t<T> CastNumpyArray(const py::object &array) {
  if (py::isinstance<py::array_t<float>>(array)) {
    return CastNumpyType<T>(array.cast<py::array_t<float>>());
  } else if (py::isinstance<py::array_t<double>>(array)) {
    return CastNumpyType<T>(array.cast<py::array_t<double>>());
  } else if (py::isinstance<py::array_t<int32_t>>(array)) {
    return CastNumpyType<T>(array.cast<py::array_t<int32_t>>());
  } else if (py::isinstance<py::array_t<int64_t>>(array)) {
    return CastNumpyType<T>(array.cast<py::array_t<int64_t>>());
  } else if (py::isinstance<py::array_t<bool>>(array)) {
    return CastNumpyType<T>(array.cast<py::array_t<bool>>());
  } else {
    PADDLE_THROW(platform::errors::InvalidArgument(
        "Value type error. The assign numpy value allows integer, float, "
        "double and bool, "
        "but received %s.",
        Py_TYPE(array.ptr())->tp_name));
  }
  // can't reach here
  return py::array_t<T>();
}

J
Jiabin Yang 已提交
410 411 412
static imperative::NameVarBaseMap ConvertToNameVarBaseMap(
    const PyNameVarBaseMap &map) {
  imperative::NameVarBaseMap result;
413 414 415 416 417 418
  for (auto &pair : map) {
    auto var_vec = GetVarBaseListFromPyHandle(pair.second);
    if (!var_vec.empty()) {
      result.emplace(pair.first, std::move(var_vec));
    }
  }
J
Jiabin Yang 已提交
419

420 421 422
  PADDLE_ENFORCE_EQ(
      PyErr_Occurred(), nullptr,
      platform::errors::InvalidArgument(py::str(py::handle(PyErr_Occurred()))));
423 424 425
  return result;
}

426 427 428 429 430 431 432 433
static bool PyCheckInteger(PyObject *obj) {
#if PY_VERSION_HEX < 0x03000000
  return (PyLong_Check(obj) || PyInt_Check(obj)) && !PyBool_Check(obj);
#else
  return PyLong_Check(obj) && !PyBool_Check(obj);
#endif
}

434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454
static Py_ssize_t GetSliceIndexFromTensor(
    const std::shared_ptr<imperative::VarBase> &tensor_index) {
  const auto &tensor = tensor_index->Var().Get<framework::LoDTensor>();
  if (tensor.numel() == 1) {
    if (tensor.type() == framework::proto::VarType::INT32) {
      return static_cast<Py_ssize_t>(operators::GetValue<int32_t>(&tensor));
    } else if (tensor.type() == framework::proto::VarType::INT64) {
      return static_cast<Py_ssize_t>(operators::GetValue<int64_t>(&tensor));
    } else {
      PADDLE_THROW(platform::errors::InvalidArgument(
          "Currently, the type of tensor in slice indices only allows "
          "int32 and int64, please check the type of index tensor."));
    }
  } else {
    PADDLE_THROW(platform::errors::InvalidArgument(
        "Currently, tensor in slice indices only allows 1 element, "
        "but received %d.",
        tensor.numel()));
  }
}

455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470
// NOTE(zhiqiu): Revised version of PySlice_GetIndices. From:
// https://github.com/python/cpython/blob/8d21aa21f2cbc6d50aab3f420bb23be1d081dac4/Objects/sliceobject.c#L103
// Original PySlice_GetIndices return wrong result when
// slice_item contains long int, such as arr[:180L].
// NOT sure why this happens !!!
// Besides, PySlice_GetIndices cannot raise error when float in slice item.
// So, I make a revised version of PySlice_GetIndices, named to
// _PySlice_GetIndices. Try to use _PySlice_Unpack which is more robust than
// PySlice_GetIndices in the future.
static int _PySlice_GetIndices(PySliceObject *r, Py_ssize_t length,
                               Py_ssize_t *start, Py_ssize_t *stop,
                               Py_ssize_t *step) {
  /* XXX support long ints */
  if (r->step == Py_None) {
    *step = 1;
  } else {
471
    if (PyCheckInteger(r->step) || IsNumpyType(r->step)) {
472
      *step = PyLong_AsLong(r->step);
473 474 475
    } else if (PyCheckTensor(r->step)) {
      *step = GetSliceIndexFromTensor(
          py::cast<std::shared_ptr<imperative::VarBase>>(r->step));
476 477
    } else {
      PADDLE_THROW(platform::errors::InvalidArgument(
478 479
          "Currently, slice indices only allows None, integers, "
          "tensor(int) and numpy(int) in slice item, but received %s.",
480 481 482 483 484 485
          std::string(Py_TYPE(r->step)->tp_name)));
    }
  }
  if (r->start == Py_None) {
    *start = *step < 0 ? length - 1 : 0;
  } else {
486
    if (PyCheckInteger(r->start) || IsNumpyType(r->start)) {
487
      *start = PyLong_AsLong(r->start);
488 489 490
    } else if (PyCheckTensor(r->start)) {
      *start = GetSliceIndexFromTensor(
          py::cast<std::shared_ptr<imperative::VarBase>>(r->start));
491 492
    } else {
      PADDLE_THROW(platform::errors::InvalidArgument(
493 494
          "Currently, slice indices only allows None, integers, "
          "tensor(int) and numpy(int) in slice item, but received %s.",
495 496 497
          std::string(Py_TYPE(r->start)->tp_name)));
    }
    if (*start < 0) *start += length;
498
    *start = std::max(*start, static_cast<Py_ssize_t>(0));
499 500 501 502
  }
  if (r->stop == Py_None) {
    *stop = *step < 0 ? -1 : length;
  } else {
503
    if (PyCheckInteger(r->stop) || IsNumpyType(r->stop)) {
504
      *stop = PyLong_AsLong(r->stop);
505 506 507
    } else if (PyCheckTensor(r->stop)) {
      *stop = GetSliceIndexFromTensor(
          py::cast<std::shared_ptr<imperative::VarBase>>(r->stop));
508 509
    } else {
      PADDLE_THROW(platform::errors::InvalidArgument(
510 511
          "Currently, slice indices only allows None, integers, "
          "tensor(int) and numpy(int) in slice item, but received %s.",
512 513
          std::string(Py_TYPE(r->stop)->tp_name)));
    }
514
    if (0 < *step && *stop < 0) *stop += length;
515
    *stop = std::min(*stop, length);
516 517 518 519 520 521 522
  }
  if (*stop > length) return -1;
  if (*start >= length) return -1;
  if (*step == 0) return -1;
  return 0;
}

Z
zyfncg 已提交
523 524 525 526 527 528 529 530 531
static void ParseIndexingSlice(
    framework::LoDTensor *tensor, PyObject *_index,
    std::vector<int> *slice_axes, std::vector<int> *slice_starts,
    std::vector<int> *slice_ends, std::vector<int> *slice_strides,
    std::vector<int> *decrease_axis, std::vector<int> *none_axes,
    std::vector<int> *infer_flags, std::vector<int> *list_select_idxs,
    bool *list_select_flag) {
  // We allow indexing by Integers, Slices, Ellipsis, None, tuples of those
  // types, and list of Bool and Integers.
S
songyouwei 已提交
532
  // wrap to tuple
533 534

  // NOTE(zhiqiu): PyTuple_Pack increases refcount.
S
songyouwei 已提交
535
  PyObject *index = !PyTuple_Check(_index) ? PyTuple_Pack(1, _index) : _index;
536 537 538 539 540 541
  DEFINE_PADDLE_SCOPE_GUARD([index, _index]() {
    if (!PyTuple_Check(_index)) {
      Py_DECREF(index);
      VLOG(4) << "Call Py_DECREF";
    }
  });
S
songyouwei 已提交
542 543 544 545 546 547
  PADDLE_ENFORCE_EQ(
      tensor->IsInitialized(), true,
      platform::errors::InvalidArgument("tensor has not been initialized"));
  const auto &shape = tensor->dims();
  const int rank = shape.size();
  const int size = PyTuple_GET_SIZE(index);
548 549 550 551

  // specified_dims is the number of dimensions which indexed by Interger,
  // Slices.
  int specified_dims = 0;
552
  int ell_count = 0;
553 554 555 556
  for (int dim = 0; dim < size; ++dim) {
    PyObject *slice_item = PyTuple_GetItem(index, dim);
    if (PyCheckInteger(slice_item) || PySlice_Check(slice_item)) {
      specified_dims++;
557 558
    } else if (slice_item == Py_Ellipsis) {
      ell_count++;
559 560 561
    }
  }

562 563 564 565
  PADDLE_ENFORCE_LE(ell_count, 1,
                    platform::errors::InvalidArgument(
                        "An index can only have a single ellipsis ('...')"));

566 567 568
  for (int i = 0, dim = 0; i < size; ++i) {
    PyObject *slice_item = PyTuple_GetItem(index, i);

S
songyouwei 已提交
569 570
    infer_flags->push_back(1);
    int dim_len = shape[dim];
571
    if (PyCheckInteger(slice_item) || IsNumpyType(slice_item)) {
572
      // integer, PyLong_AsLong supports both int and long
S
songyouwei 已提交
573
      int start = static_cast<int>(PyLong_AsLong(slice_item));
H
hong 已提交
574
      auto s_t = start;
S
songyouwei 已提交
575
      start = start < 0 ? start + dim_len : start;
576
      if (start >= dim_len || start < 0) {
H
hong 已提交
577 578 579 580 581 582 583 584 585
        std::string str_error_message =
            "The starting index " + std::to_string(s_t) +
            " of slice is out of bounds in tensor " + std::to_string(dim) +
            "-th axis, it shound be in the range of [" +
            std::to_string(-dim_len) + ", " + std::to_string(dim_len) + ")";
        // py::index_error is corresponding to IndexError in Python
        // Used to indicate out of bounds access in __getitem__, __setitem__
        throw py::index_error(str_error_message);
      }
S
songyouwei 已提交
586 587 588 589 590
      slice_axes->push_back(dim);
      slice_starts->push_back(start);
      slice_ends->push_back(start + 1);
      slice_strides->push_back(1);
      decrease_axis->push_back(dim);
591 592
      dim++;
    } else if (PySlice_Check(slice_item)) {
593
      // slice item
S
songyouwei 已提交
594
      Py_ssize_t start, end, step;
595 596 597
      PySliceObject *p = reinterpret_cast<PySliceObject *>(slice_item);
      _PySlice_GetIndices(p, dim_len, &start, &end, &step);

S
songyouwei 已提交
598
      // :: or : or 0:dim_len:1
599
      if (start == 0 && end == dim_len && step == 1) {
600
        dim++;
601 602
        continue;
      }
S
songyouwei 已提交
603 604 605 606
      slice_axes->push_back(dim);
      slice_starts->push_back(start);
      slice_ends->push_back(end);
      slice_strides->push_back(step);
607 608 609
      dim++;
    } else if (slice_item == Py_Ellipsis) {
      dim += rank - specified_dims;
610 611
    } else if (slice_item == Py_None) {
      none_axes->push_back(dim);
Z
zyfncg 已提交
612 613
    } else if (PyList_Check(slice_item)) {
      *list_select_flag = true;
Z
zyfncg 已提交
614 615 616 617 618 619
      PADDLE_ENFORCE_EQ(
          size, 1,
          platform::errors::InvalidArgument(
              "When index contains a list, its length is excepted to 1, "
              "but received %d",
              size));
Z
zyfncg 已提交
620 621 622 623 624 625 626 627 628 629 630 631
      bool all_bool = true;
      int list_size = PyList_GET_SIZE(slice_item);
      for (int j = 0; j < list_size; ++j) {
        PyObject *list_item = PyList_GetItem(slice_item, j);
        if (PyCheckInteger(list_item)) {
          all_bool = false;
        } else if (!PyBool_Check(list_item)) {
          PADDLE_THROW(platform::errors::InvalidArgument(
              "Only support int or bool in index list."));
        }
      }
      if (all_bool) {
Z
zyfncg 已提交
632 633 634 635 636 637 638
        PADDLE_ENFORCE_EQ(
            list_size, shape[0],
            platform::errors::InvalidArgument(
                "The dimension of bool index doesn't match indexed array along "
                "dimension 0, the target dimension is %d, but received %d.",
                shape[0], list_size));

Z
zyfncg 已提交
639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658
        for (int j = 0; j < list_size; ++j) {
          PyObject *list_item = PyList_GetItem(slice_item, j);
          if (list_item == Py_True) {
            list_select_idxs->push_back(j);
          }
        }
      } else {
        for (int j = 0; j < list_size; ++j) {
          PyObject *list_item = PyList_GetItem(slice_item, j);
          if (PyCheckInteger(list_item)) {
            list_select_idxs->push_back(
                static_cast<int>(PyLong_AsLong(list_item)));
          } else if (list_item == Py_True) {
            list_select_idxs->push_back(1);
          } else {
            list_select_idxs->push_back(0);
          }
        }
      }

659 660
    } else {
      PADDLE_THROW(platform::errors::InvalidArgument(
661
          "Currently, Tensor.__indices__() only allows indexing "
Z
zyfncg 已提交
662 663 664
          "by Integers, Slices, Ellipsis, None, tuples of these types "
          "and list of Bool and Integers, but received "
          "%s in %dth slice item",
665
          std::string(Py_TYPE(slice_item)->tp_name), i + 1));
S
songyouwei 已提交
666 667
    }
  }
668 669

  // valid_index is the number of dimensions exclude None index
670
  const int valid_indexs = size - none_axes->size() - ell_count;
671 672 673 674
  PADDLE_ENFORCE_EQ(valid_indexs <= rank, true,
                    platform::errors::InvalidArgument(
                        "Too many indices (%d) for tensor of dimension %d.",
                        valid_indexs, rank));
S
songyouwei 已提交
675 676
}

677
template <typename P>
678 679 680
static void VarBaseCopy(std::shared_ptr<imperative::VarBase> &src,  // NOLINT
                        imperative::VarBase &dst,                   // NOLINT
                        const P &dst_device, const bool blocking) {
681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732
  if (dst.SharedVar()->IsEmpty()) {
    VLOG(3) << "deep copy Variable from " << src->Name() << " to "
            << dst.Name();
    dst.SetPersistable(src->Persistable());
    dst.SetDataType(src->DataType());
    dst.SetType(src->Type());
    dst.SetOverridedStopGradient(src->OverridedStopGradient());
    if (!src->SharedVar()->IsEmpty()) {
      if (src->Var().IsType<framework::LoDTensor>()) {
        auto &src_tensor = src->Var().Get<framework::LoDTensor>();
        auto *dst_tensor = dst.MutableVar()->GetMutable<framework::LoDTensor>();
        dst_tensor->set_lod(src_tensor.lod());
        framework::TensorCopy(src_tensor, dst_device, dst_tensor);
        if (blocking) {
          platform::DeviceContextPool::Instance().Get(dst_device)->Wait();
          auto src_device = src_tensor.place();
          if (!(src_device == dst_device)) {
            platform::DeviceContextPool::Instance().Get(src_device)->Wait();
          }
        }
      } else if (src->Var().IsType<framework::SelectedRows>()) {
        auto &src_selected_rows = src->Var().Get<framework::SelectedRows>();
        auto *dst_selected_rows =
            dst.MutableVar()->GetMutable<framework::SelectedRows>();
        dst_selected_rows->set_height(src_selected_rows.height());
        dst_selected_rows->set_rows(src_selected_rows.rows());
        framework::TensorCopy(src_selected_rows.value(), dst_device,
                              dst_selected_rows->mutable_value());
        if (blocking) {
          platform::DeviceContextPool::Instance().Get(dst_device)->Wait();
          auto src_device = src_selected_rows.value().place();
          if (!(src_device == dst_device)) {
            platform::DeviceContextPool::Instance().Get(src_device)->Wait();
          }
        }
      }

      if (!blocking) {
        IncreaseVarbaseReferenceCountUntilCopyComplete(src, dst_device);
      }

    } else {
      PADDLE_THROW(platform::errors::InvalidArgument(
          "The source Tensor(%s) can not copy when it is empty.", src->Name()));
    }
  } else {
    PADDLE_THROW(platform::errors::InvalidArgument(
        "The destion Tensor(%s) can not copy when it is not empty.",
        dst.Name()));
  }
}

733
// Bind Methods
J
Jiabin Yang 已提交
734
void BindImperative(py::module *m_ptr) {
735 736
  auto &m = *m_ptr;

737 738
  BindOpFunctions(&m);

739 740
#ifndef _WIN32
  // Dygraph DataLoader signal handler
741 742 743 744 745 746 747 748 749 750 751 752 753
  m.def("_set_process_pids", [](int64_t key, py::object &obj) {
    PADDLE_ENFORCE_EQ(
        py::isinstance<py::tuple>(obj) || py::isinstance<py::list>(obj), true,
        platform::errors::InvalidArgument(
            "The subprocess ids set in DataLoader is illegal."
            "Expected data type is tuple or list, but received %s",
            obj.get_type()));
    py::list pids = py::cast<py::list>(obj);
    std::set<pid_t> pids_set = {};
    for (size_t i = 0; i < pids.size(); i++) {
      pids_set.insert(pids[i].cast<pid_t>());
    }
    imperative::SetLoadProcessPIDs(key, pids_set);
754
  });
755 756
  m.def("_erase_process_pids",
        [](int64_t key) { imperative::EraseLoadProcessPIDs(key); });
757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808
  m.def("_set_process_signal_handler",
        []() { imperative::SetLoadProcessSignalHandler(); });
  m.def("_throw_error_if_process_failed",
        []() { imperative::ThrowErrorIfLoadProcessFailed(); });

  // Dygraph DataLoader reader process & thread related functions
  m.def(
      "_convert_to_tensor_list",
      [](py::object &obj) -> py::list {
        // 0. input data check
        PADDLE_ENFORCE(
            py::isinstance<py::tuple>(obj) || py::isinstance<py::list>(obj),
            platform::errors::InvalidArgument(
                "The batch data read into DataLoader is illegal."
                "Expected data type is tuple or list, but received %s",
                obj.get_type()));
        py::list batch = py::cast<py::list>(obj);
        py::list tensors;
        for (size_t i = 0; i < batch.size(); ++i) {
          // 1. cast to python array
          auto array = batch[i].cast<py::array>();
          PADDLE_ENFORCE_NE(
              string::Sprintf("%s", array.dtype()).compare("object"), 0,
              platform::errors::InvalidArgument(
                  "Faild to convert input data to a regular ndarray.\n  * "
                  "Usually this means the input data contains nested "
                  "lists with different lengths.\n  * Check the reader "
                  "function passed to 'set_(sample/sample_list/batch)"
                  "_generator' to locate the data causes this issue."));
          // 2. construcct LoDTensor
          framework::LoDTensor t;
          SetTensorFromPyArray<platform::CPUPlace>(&t, array,
                                                   platform::CPUPlace(), true);
          // 3. allocate shared memory
          void *data_ptr = t.data<void>();
          size_t data_size = t.numel() * framework::SizeOfType(t.type());
          auto shared_writer_holder =
              memory::allocation::AllocateMemoryMapWriterAllocation(data_size);
          // 4. maintain mmap fd set & backup ipc_name
          const std::string &ipc_name = shared_writer_holder->ipc_name();
          memory::allocation::MemoryMapFdSet::Instance().Insert(ipc_name);
          // 5. copy data & reset holder
          memory::Copy(platform::CPUPlace(), shared_writer_holder->ptr(),
                       platform::CPUPlace(), data_ptr, data_size);
          t.ResetHolder(shared_writer_holder);
          // 6. append to result list
          tensors.append(t);
        }
        return tensors;
      },
      py::return_value_policy::take_ownership);

K
Kaipeng Deng 已提交
809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841
  m.def("_array_to_share_memory_tensor",
        [](py::object &obj) {
          // 1. cast to python array
          auto array = obj.cast<py::array>();
          PADDLE_ENFORCE_NE(
              string::Sprintf("%s", array.dtype()).compare("object"), 0,
              platform::errors::InvalidArgument(
                  "Faild to convert input data to a regular ndarray.\n  * "
                  "Usually this means the input data contains nested "
                  "lists with different lengths.\n  * Check the reader "
                  "function passed to 'set_(sample/sample_list/batch)"
                  "_generator' to locate the data causes this issue."));
          // 2. construcct LoDTensor
          framework::LoDTensor t;
          SetTensorFromPyArray<platform::CPUPlace>(&t, array,
                                                   platform::CPUPlace(), true);
          // 3. allocate shared memory
          void *data_ptr = t.data<void>();
          size_t data_size = t.numel() * framework::SizeOfType(t.type());
          auto shared_writer_holder =
              memory::allocation::AllocateMemoryMapWriterAllocation(data_size);
          // 4. maintain mmap fd set & backup ipc_name
          const std::string &ipc_name = shared_writer_holder->ipc_name();
          memory::allocation::MemoryMapFdSet::Instance().Insert(ipc_name);
          // 5. copy data & reset holder
          memory::Copy(platform::CPUPlace(), shared_writer_holder->ptr(),
                       platform::CPUPlace(), data_ptr, data_size);
          t.ResetHolder(shared_writer_holder);

          return t;
        },
        py::return_value_policy::take_ownership);

842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861
  m.def("_remove_tensor_list_mmap_fds", [](py::list &tensor_list) {
    for (size_t i = 0; i < tensor_list.size(); ++i) {
      auto t = tensor_list[i].cast<framework::LoDTensor>();
      auto *mmap_writer_allocation =
          dynamic_cast<memory::allocation::MemoryMapWriterAllocation *>(
              t.Holder().get());
      PADDLE_ENFORCE_NOT_NULL(
          mmap_writer_allocation,
          platform::errors::NotFound("The shared memory of LoDTensor in "
                                     "DataLoader's child process has been "
                                     "released."));
      memory::allocation::MemoryMapFdSet::Instance().Remove(
          mmap_writer_allocation->ipc_name());
    }
  });

  m.def("_cleanup_mmap_fds",
        []() { memory::allocation::MemoryMapFdSet::Instance().Clear(); });
#endif

862 863 864 865 866
  m.def("start_imperative_gperf_profiler",
        []() { imperative::StartProfile(); });

  m.def("stop_imperative_gperf_profiler", []() { imperative::StopProfile(); });

Z
Zeng Jinle 已提交
867 868 869
  m.def("_is_dygraph_debug_enabled",
        []() { return imperative::IsDebugEnabled(); });
  m.def("_dygraph_debug_level", []() { return imperative::GetDebugLevel(); });
870 871 872 873
  m.def("_switch_tracer",
        [](const std::shared_ptr<imperative::Tracer> &tracer) {
          imperative::SetCurrentTracer(tracer);
        });
Z
Zeng Jinle 已提交
874

875 876 877 878
  py::class_<imperative::VarBase, std::shared_ptr<imperative::VarBase>> varbase(
      m, "VarBase", R"DOC()DOC");
  g_varbase_pytype = (PyTypeObject *)varbase.ptr();  // NOLINT
  varbase.def_static("_alive_vars", &imperative::VarBase::AliveVarNames)
879 880 881 882 883 884 885
      .def("__init__",
           [](imperative::VarBase &self) {
             std::string name =
                 imperative::GetCurrentTracer()->GenerateUniqueName(
                     "generated_tensor");
             new (&self) imperative::VarBase(name);
           })
J
Jiabin Yang 已提交
886
      .def("__init__",
887 888 889
           [](imperative::VarBase &self, framework::proto::VarType::Type dtype,
              const std::vector<int> &dims, const py::handle &name,
              framework::proto::VarType::Type type, bool persistable) {
890
             VLOG(4) << "Init VarBase";
891 892 893
             std::string act_name = "";
             if (!name.ptr() || name.ptr() == Py_None) {
               act_name = imperative::GetCurrentTracer()->GenerateUniqueName(
894
                   "generated_tensor");
895 896 897 898
             } else {
               act_name = name.cast<std::string>();
             }
             new (&self) imperative::VarBase(act_name);
J
Jiabin Yang 已提交
899 900 901 902 903 904 905 906 907
             self.SetPersistable(persistable);
             self.SetType(type);
             self.SetDataType(dtype);
             if (type == framework::proto::VarType::LOD_TENSOR) {
               auto *tensor =
                   self.MutableVar()->GetMutable<framework::LoDTensor>();
               tensor->Resize(framework::make_ddim(dims));
             }
           })
908 909
      .def("__init__", &InitVarBaseFromNumpyWithArg<platform::CPUPlace>,
           py::arg("value"), py::arg("place"), py::arg("persistable") = false,
910 911
           py::arg("zero_copy") = false, py::arg("name") = "",
           py::arg("stop_gradient") = -1)
912 913 914 915
      .def("__init__", &InitVarBaseFromNumpyWithArg<platform::XPUPlace>,
           py::arg("value"), py::arg("place"), py::arg("persistable") = false,
           py::arg("zero_copy") = false, py::arg("name") = "",
           py::arg("stop_gradient") = -1)
916 917
      .def("__init__", &InitVarBaseFromNumpyWithArg<platform::CUDAPlace>,
           py::arg("value"), py::arg("place"), py::arg("persistable") = false,
918 919
           py::arg("zero_copy") = false, py::arg("name") = "",
           py::arg("stop_gradient") = -1)
920 921
      .def("__init__", &InitVarBaseFromNumpyWithArg<platform::CUDAPinnedPlace>,
           py::arg("value"), py::arg("place"), py::arg("persistable") = false,
922 923
           py::arg("zero_copy") = false, py::arg("name") = "",
           py::arg("stop_gradient") = -1)
924 925 926 927
      .def("__init__", &InitVarBaseFromNumpyWithArg<platform::NPUPlace>,
           py::arg("value"), py::arg("place"), py::arg("persistable") = false,
           py::arg("zero_copy") = false, py::arg("name") = "",
           py::arg("stop_gradient") = -1)
L
Leo Chen 已提交
928
      .def("__init__", &InitVarBaseFromNumpyWithArgDefault, py::arg("value"))
929
      .def("__init__", &InitVarBaseFromTensorWithArgDefault, py::arg("tensor"))
930 931 932 933 934 935 936 937 938 939
      .def("__init__", &InitVarBaseFromTensorWithArg<platform::CPUPlace>,
           py::arg("tensor"), py::arg("place"))
      .def("__init__", &InitVarBaseFromTensorWithArg<platform::XPUPlace>,
           py::arg("tensor"), py::arg("place"))
      .def("__init__", &InitVarBaseFromTensorWithArg<platform::CUDAPlace>,
           py::arg("tensor"), py::arg("place"))
      .def("__init__", &InitVarBaseFromTensorWithArg<platform::CUDAPinnedPlace>,
           py::arg("tensor"), py::arg("place"))
      .def("__init__", &InitVarBaseFromTensorWithArg<platform::NPUPlace>,
           py::arg("tensor"), py::arg("place"))
940
      .def("__init__", &InitVarBaseFromNumpyWithKwargs)
941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025
      .def(
          "__setitem_varbase__",
          [](std::shared_ptr<imperative::VarBase> &self, py::handle _index,
             py::object &value_obj) {
            VLOG(4) << "Call __setitem_varbase__";

            auto self_tensor =
                self->MutableVar()->GetMutable<framework::LoDTensor>();
            // NOTE(zhiqiu): PyTuple_Pack increases refcount while PyTuple_New
            // https://github.com/python/cpython/blob/24b63c695ae0a95b06379eaadace66735abac1e2/Objects/tupleobject.c#L251
            PyObject *index_ptr = !PyTuple_Check(_index.ptr())
                                      ? PyTuple_Pack(1, _index.ptr())
                                      : _index.ptr();
            DEFINE_PADDLE_SCOPE_GUARD([index_ptr, &_index]() {
              if (!PyTuple_Check(_index.ptr())) {
                Py_DECREF(index_ptr);
                VLOG(4) << "Call Py_DECREF";
              }
            });

            auto is_tensor = [](py::handle var) {
              if (!var.ptr() || var.ptr() == Py_None) {
                return false;
              }
              try {
                py::cast<std::shared_ptr<imperative::VarBase>>(var);
                return true;
              } catch (py::cast_error &) {
                return false;
              }
            };

            // 1. Check argumnets
            bool parse_index = true;

            // Check whether _index can be parsed.
            const int size = PyTuple_GET_SIZE(index_ptr);
            for (int dim = 0; dim < size; ++dim) {
              PyObject *slice_item = PyTuple_GetItem(index_ptr, dim);
              if (!(PyCheckInteger(slice_item) || PySlice_Check(slice_item) ||
                    slice_item == Py_Ellipsis || slice_item == Py_None)) {
                parse_index = false;
                break;
              }
            }

            // 2. Call op set_value to speed up if the condition is met,
            // otherwise call TensorToPyArray.
            // TODO(liym27): Try not to call TensorToPyArray because it always
            // copys data to cpu place, which reduces performance.
            if (parse_index) {
              std::vector<int> axes, starts, ends, steps, decrease_axes,
                  none_axes, infer_flags, list_select_idxs;
              // if index is a list, list_select_flag will be true
              bool list_select_flag = false;
              ParseIndexingSlice(self_tensor, index_ptr, &axes, &starts, &ends,
                                 &steps, &decrease_axes, &none_axes,
                                 &infer_flags, &list_select_idxs,
                                 &list_select_flag);

              framework::AttributeMap attrs = {{"axes", axes},
                                               {"starts", starts},
                                               {"ends", ends},
                                               {"steps", steps},
                                               {"decrease_axes", decrease_axes},
                                               {"none_axes", none_axes}};

              imperative::NameVarBaseMap ins = {{"Input", {self}}};
              imperative::NameVarBaseMap outs = {{"Out", {self}}};

              const auto &tracer = imperative::GetCurrentTracer();

              if (tracer->HasGrad()) {
                PADDLE_ENFORCE_EQ(
                    self->IsLeaf() && !self->OverridedStopGradient(), false,
                    platform::errors::InvalidArgument(
                        "Leaf Tensor (%s) that doesn't stop gradient can't use "
                        "inplace strategy.",
                        self->Name()));
              }

              if (PyCheckTensor(value_obj.ptr())) {
                auto value_tensor =
                    value_obj.cast<std::shared_ptr<imperative::VarBase>>();
                ins.insert({"ValueTensor", {value_tensor}});
1026 1027 1028 1029 1030 1031

                // pass the stop_gradient from value to tensor
                if (!value_tensor->OverridedStopGradient() &&
                    self->OverridedStopGradient()) {
                  self->SetOverridedStopGradient(false);
                }
1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144
              } else if (py::isinstance<py::array>(value_obj)) {
                auto value_tensor = std::shared_ptr<imperative::VarBase>(
                    new imperative::VarBase(false,
                                            tracer->GenerateUniqueName()));
                py::object value = value_obj;
                if (self->DataType() == framework::proto::VarType::FP32) {
                  if (!py::isinstance<py::array_t<float>>(value_obj)) {
                    value = CastNumpyArray<float>(value_obj);
                  }
                } else if (self->DataType() ==
                           framework::proto::VarType::FP64) {
                  if (!py::isinstance<py::array_t<double>>(value_obj)) {
                    value = CastNumpyArray<double>(value_obj);
                  }
                } else if (self->DataType() ==
                           framework::proto::VarType::INT32) {
                  if (!py::isinstance<py::array_t<int32_t>>(value_obj)) {
                    value = CastNumpyArray<int32_t>(value_obj);
                  }
                } else if (self->DataType() ==
                           framework::proto::VarType::INT64) {
                  if (!py::isinstance<py::array_t<int64_t>>(value_obj)) {
                    value = CastNumpyArray<int64_t>(value_obj);
                  }
                } else if (self->DataType() ==
                           framework::proto::VarType::BOOL) {
                  if (!py::isinstance<py::array_t<bool>>(value_obj)) {
                    value = CastNumpyArray<bool>(value_obj);
                  }
                } else {
                  PADDLE_THROW(platform::errors::InvalidArgument(
                      "When assign a numpy.np value to a paddle.Tensor, "
                      "the data type of the paddle.Tensor must be bool, "
                      "float32, int32 or int64, "
                      "please check the type of tensor."));
                }

                SetTensorFromPyArray(value_tensor->MutableVar()
                                         ->GetMutable<framework::LoDTensor>(),
                                     value, self->Place(), false);
                ins.insert({"ValueTensor", {value_tensor}});

              } else {
                // convert the value to self data type
                if (py::isinstance<py::float_>(value_obj) ||
                    py::isinstance<py::int_>(value_obj) ||
                    py::isinstance<py::bool_>(value_obj)) {
                  if (self->DataType() == framework::proto::VarType::FP32) {
                    attrs["fp32_values"] =
                        std::vector<float>{value_obj.cast<float>()};
                  } else if (self->DataType() ==
                             framework::proto::VarType::FP64) {
                    attrs["fp64_values"] =
                        std::vector<double>{value_obj.cast<double>()};
                  } else if (self->DataType() ==
                             framework::proto::VarType::INT32) {
                    attrs["int32_values"] =
                        std::vector<int32_t>{value_obj.cast<int32_t>()};
                  } else if (self->DataType() ==
                             framework::proto::VarType::INT64) {
                    attrs["int64_values"] =
                        std::vector<int64_t>{value_obj.cast<int64_t>()};
                  } else if (self->DataType() ==
                             framework::proto::VarType::BOOL) {
                    attrs["bool_values"] =
                        std::vector<int>{value_obj.cast<bool>()};
                  } else {
                    PADDLE_THROW(platform::errors::InvalidArgument(
                        "When assign a value to a paddle.Tensor, "
                        "the data type of the paddle.Tensor must be bool, "
                        "float32, int32 or int64, "
                        "please check the type of tensor."));
                  }
                  attrs["shape"] = std::vector<int64_t>{1};

                } else {
                  PADDLE_THROW(platform::errors::InvalidArgument(
                      "Value type error. The assign value allows "
                      "numpy.ndarray, integer, float or bool, "
                      "but received %s.",
                      Py_TYPE(value_obj.ptr())));
                }
              }

              {
                // Release gil and do tracing
                py::gil_scoped_release release;
                tracer->TraceOp("set_value", ins, outs, std::move(attrs),
                                {{"Input", "Out"}});
              }
            } else {
              auto self_numpy = TensorToPyArray(*self_tensor);
              VLOG(4) << "parse_index is false";
              if (is_tensor(_index)) {
                VLOG(4) << "index is tensor";
                auto index_var =
                    py::cast<std::shared_ptr<imperative::VarBase>>(_index);
                auto index_tensor =
                    index_var->MutableVar()->GetMutable<framework::LoDTensor>();
                auto index_numpy = TensorToPyArray(*index_tensor);
                self_numpy[index_numpy] = value_obj;
              } else {
                VLOG(4) << "index is not tensor";
                self_numpy[_index] = value_obj;
              }
              SetTensorFromPyArray(self_tensor, self_numpy,
                                   self_tensor->place(), false);
            }
            // NOTE(liym27):
            // Increase the version of VarBase self because __setitem__ is an
            // inplace operator for the VarBase self.
            self->BumpInplaceVersion();
          })
1145
      .def("_getitem_index_not_tensor",
S
songyouwei 已提交
1146
           [](std::shared_ptr<imperative::VarBase> &self, py::handle _index) {
1147
             VLOG(4) << "Call _getitem_index_not_tensor";
1148
             std::vector<int> slice_axes, slice_starts, slice_ends,
Z
zyfncg 已提交
1149 1150 1151 1152
                 slice_strides, decrease_axis, none_axes, infer_flags,
                 list_select_idxs;
             // if index is a list, list_select_flag will be true
             bool list_select_flag = false;
S
songyouwei 已提交
1153 1154 1155 1156
             auto tensor =
                 self->MutableVar()->GetMutable<framework::LoDTensor>();
             ParseIndexingSlice(tensor, _index.ptr(), &slice_axes,
                                &slice_starts, &slice_ends, &slice_strides,
Z
zyfncg 已提交
1157 1158
                                &decrease_axis, &none_axes, &infer_flags,
                                &list_select_idxs, &list_select_flag);
1159 1160 1161
             // release gil and do tracing
             py::gil_scoped_release release;
             const auto &tracer = imperative::GetCurrentTracer();
1162

Z
zyfncg 已提交
1163
             auto out = slice_axes.empty() && !list_select_flag
1164 1165 1166 1167
                            ? self
                            : std::shared_ptr<imperative::VarBase>(
                                  new imperative::VarBase(
                                      tracer->GenerateUniqueName()));
Z
zyfncg 已提交
1168

1169
             if (!slice_axes.empty()) {
S
songyouwei 已提交
1170
               imperative::NameVarBaseMap ins = {{"Input", {self}}};
1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188
               framework::AttributeMap attrs = {
                   {"axes", slice_axes},
                   {"starts", slice_starts},
                   {"ends", slice_ends},
                   {"infer_flags", infer_flags},
                   {"decrease_axis", decrease_axis}};
               imperative::NameVarBaseMap outs = {{"Out", {out}}};
               std::string op_type = "slice";
               for (auto stride : slice_strides) {
                 if (stride != 1) {
                   op_type = "strided_slice";
                   attrs.insert({"strides", slice_strides});
                   attrs.erase("decrease_axis");
                   break;
                 }
               }
               tracer->TraceOp(op_type, ins, outs, std::move(attrs));
             }
1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216
             if (!none_axes.empty()) {
               // Deal with cases when all axes are decreased.
               // After slice, the shape of out is [1], which should have been
               // [], but Paddle doesn't support scalar.
               // In order to ensure the correctness of the final shape of out,
               // one dimension of out needs to be decreased.
               // For example:
               // # x.shape: (2,3,4)
               // out = x[0, 1, 1, None] # out.shape : (1)
               if (static_cast<int>(decrease_axis.size()) ==
                   tensor->dims().size()) {
                 none_axes.pop_back();
               }
               if (!none_axes.empty()) {
                 // Deal with cases that decrease_axes is not empty
                 // For example:
                 // # x.shape: (2,3,4)
                 // out = x[0, 0:2, None] # out.shape : (2, 1, 4)
                 for (auto &axis : none_axes) {
                   int len = 0;
                   for (int da : decrease_axis) {
                     if (da < axis) {
                       len++;
                     }
                   }
                   axis -= len;
                 }

1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239
                 // Deal with cases that there are more than one
                 // prefix none index, For example:
                 // [None, None, :, :, None]
                 // the none_axes int the return of ParseIndexingSlice is:
                 // [0,    0,          2   ]
                 // according to the interface of "unsqueeze2",
                 // we should convert it to:
                 // [0,    0,          4   ]
                 int prefix_zero_cnt = 0;
                 for (const auto &axis : none_axes) {
                   if (axis == 0) {
                     prefix_zero_cnt++;
                   } else {
                     break;
                   }
                 }
                 if (prefix_zero_cnt > 0) {
                   int none_axes_num = static_cast<int>(none_axes.size());
                   for (int i = prefix_zero_cnt; i < none_axes_num; ++i) {
                     none_axes[i] += prefix_zero_cnt;
                   }
                 }

1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253
                 imperative::NameVarBaseMap ins = {{"X", {out}}};
                 framework::AttributeMap attrs = {{"axes", none_axes}};
                 auto new_out = std::shared_ptr<imperative::VarBase>(
                     new imperative::VarBase(tracer->GenerateUniqueName()));
                 auto out_xshape = std::shared_ptr<imperative::VarBase>(
                     new imperative::VarBase(tracer->GenerateUniqueName()));
                 imperative::NameVarBaseMap outs = {{"Out", {new_out}},
                                                    {"XShape", {out_xshape}}};
                 tracer->TraceOp("unsqueeze2", ins, outs, std::move(attrs));

                 return new_out;
               }
             }

Z
zyfncg 已提交
1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269
             // the index is a list
             if (list_select_flag) {
               auto select_index = std::shared_ptr<imperative::VarBase>(
                   new imperative::VarBase(tracer->GenerateUniqueName()));
               auto *idx_tensor = select_index->MutableVar()
                                      ->GetMutable<framework::LoDTensor>();
               auto *dev_ctx = platform::DeviceContextPool::Instance().Get(
                   tracer->ExpectedPlace());
               TensorFromVector(list_select_idxs, *dev_ctx, idx_tensor);

               imperative::NameVarBaseMap ins = {{"X", {self}},
                                                 {"Index", {select_index}}};
               imperative::NameVarBaseMap outs = {{"Out", {out}}};
               tracer->TraceOp("index_select", ins, outs, {{"dim", 0}});
             }

1270
             return out;
1271
           })
1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335
      .def(
          "_getitem_from_offset",
          [](std::shared_ptr<imperative::VarBase> &self, const py::args &args) {
            const auto &tensor = self->Var().Get<framework::LoDTensor>();
            PADDLE_ENFORCE_EQ(
                tensor.IsInitialized(), true,
                platform::errors::InvalidArgument(
                    "Tensor of %s is Empty, please check if it has no data.",
                    self->Name()));

            const auto &tensor_dims = tensor.dims();

            std::vector<size_t> dims(tensor_dims.size());
            std::vector<size_t> strides(tensor_dims.size());

            size_t numel = 1;
            for (int i = tensor_dims.size() - 1; i >= 0; --i) {
              strides[i] = numel;
              dims[i] = static_cast<size_t>(tensor_dims[i]);
              numel *= dims[i];
            }
            size_t offset = 0;
            if (args.empty()) {
              PADDLE_ENFORCE_EQ(
                  numel, 1,
                  platform::errors::InvalidArgument(
                      "only one element tensors can be converted to Python "
                      "scalars when no input coordinates"));
            } else if (args.size() == 1) {
              offset = args[0].cast<size_t>();
              PADDLE_ENFORCE_LT(
                  offset, numel,
                  platform::errors::InvalidArgument(
                      "index %d is out of bounds for size %d", offset, numel));
            } else {
              PADDLE_ENFORCE_EQ(args.size(), dims.size(),
                                platform::errors::InvalidArgument(
                                    "incorrect number of indices for Tensor"));

              for (size_t i = 0; i < args.size(); ++i) {
                size_t index = args[i].cast<size_t>();
                PADDLE_ENFORCE_LT(
                    index, dims[i],
                    platform::errors::InvalidArgument(
                        "index %d is out fo bounds for axis %d with size %d",
                        index, i, dims[i]));
                offset += index * strides[i];
              }
            }
#define TENSOR_TO_PY_SCALAR(T, proto_type)                                   \
  if (tensor.type() == proto_type) {                                         \
    std::string py_dtype_str = details::TensorDTypeToPyDTypeStr(proto_type); \
    T b = TensorGetElement<T>(tensor, offset);                               \
    return py::array(py::dtype(py_dtype_str.c_str()), {}, {},                \
                     static_cast<void *>(&b));                               \
  }

            _ForEachDataType_(TENSOR_TO_PY_SCALAR);
#undef TENSOR_TO_PY_SCALAR
            PADDLE_THROW(platform::errors::Unimplemented(
                "Unsupported tensor data type: %s",
                framework::DataTypeToString(tensor.type())));
          },
          py::return_value_policy::copy)
1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357
      .def("_inplace_version",
           [](imperative::VarBase &self) -> uint32_t {
             const auto &var = self.MutableVar();
             PADDLE_ENFORCE_EQ(
                 var->IsInitialized(), true,
                 platform::errors::InvalidArgument(
                     "Tensor of %s is Empty, please check if it has no data.",
                     self.Name()));
             return var->CurrentInplaceVersion();
           })
      .def("_bump_inplace_version",
           [](std::shared_ptr<imperative::VarBase> &self) {
             // NOTE(liym27): _bump_inplace_version is only used for inplace
             // operation
             self->BumpInplaceVersion();
           },
           R"DOC(
        **Notes**:
            **This API is ONLY available in Dygraph mode.**
            **This is a very low level API. Users should not use it directly. **
         Bump the version whenever the Tensor is modified through an inplace operation.
            )DOC")
1358
      .def("numpy",
1359

1360 1361 1362 1363 1364 1365
           [](imperative::VarBase &self) -> py::array {
             const auto &tensor =
                 self.MutableVar()->Get<framework::LoDTensor>();
             PADDLE_ENFORCE_EQ(
                 tensor.IsInitialized(), true,
                 platform::errors::InvalidArgument(
1366
                     "Tensor of %s is Empty, please check if it has no data.",
1367 1368 1369 1370
                     self.Name()));
             return TensorToPyArray(tensor, true);
           },
           R"DOC(
Z
Zhou Wei 已提交
1371 1372
        Returns a numpy array shows the value of current Tensor.
        
1373
        Returns:
Z
Zhou Wei 已提交
1374
            ndarray: The numpy value of current Tensor.
1375 1376

        Returns type:
Z
Zhou Wei 已提交
1377
            ndarray: dtype is same as current Tensor
1378 1379 1380 1381

        Examples:
            .. code-block:: python

Z
Zhou Wei 已提交
1382
                import paddle
1383 1384
                import numpy as np
                data = np.random.uniform(-1, 1, [30, 10, 32]).astype('float32')
Z
Zhou Wei 已提交
1385 1386 1387 1388
                linear = paddle.nn.Linear(32, 64)
                data = paddle.to_tensor(data)
                x = linear(data)
                print(x.numpy())
1389
       )DOC")
1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452
      .def("detach",
           [](const imperative::VarBase
                  &self) -> std::shared_ptr<imperative::VarBase> {
             PADDLE_ENFORCE_EQ(
                 self.Var().IsInitialized(), true,
                 platform::errors::InvalidArgument(
                     "Tensor %s has not been initialized!", self.Name()));

             PADDLE_ENFORCE_EQ(
                 self.Var().IsType<framework::LoDTensor>() ||
                     self.Var().IsType<framework::SelectedRows>(),
                 true,
                 platform::errors::InvalidArgument(
                     "Type of Tensor[%s] must be LoDTensor or SelectedRows!",
                     self.Name()));

             auto detach_var = std::make_shared<imperative::VarBase>(
                 true, "detach_" + self.Name());

             detach_var->SetPersistable(self.Persistable());
             detach_var->SetType(self.Type());
             detach_var->SetDataType(self.DataType());

             if (self.Var().IsType<framework::LoDTensor>()) {
               const auto &origin_tensor =
                   self.Var().Get<framework::LoDTensor>();
               PADDLE_ENFORCE_EQ(
                   origin_tensor.IsInitialized(), true,
                   platform::errors::InvalidArgument(
                       "Tensor %s has not been initialized!", self.Name()));

               auto *detach_tensor =
                   detach_var->MutableVar()->GetMutable<framework::LoDTensor>();
               detach_tensor->ShareDataWith(origin_tensor);
               // NOTE(liym27): Call ShareInplaceVersionCounterWith to share the
               // same TensorInplaceVersion, which is used to check whether
               // inplace
               // operations are correct.
               detach_tensor->ShareInplaceVersionCounterWith(origin_tensor);
             } else {
               const auto &origin_selected_rows =
                   self.Var().Get<framework::SelectedRows>();
               PADDLE_ENFORCE_EQ(
                   origin_selected_rows.value().IsInitialized(), true,
                   platform::errors::InvalidArgument(
                       "Tensor %s has not been initialized!", self.Name()));

               auto *detach_selected_rows =
                   detach_var->MutableVar()
                       ->GetMutable<framework::SelectedRows>();
               detach_selected_rows->set_height(origin_selected_rows.height());
               detach_selected_rows->set_rows(origin_selected_rows.rows());
               detach_selected_rows->mutable_value()->ShareDataWith(
                   origin_selected_rows.value());
               detach_selected_rows->mutable_value()
                   ->ShareInplaceVersionCounterWith(
                       origin_selected_rows.value());
             }
             VLOG(3) << "The detached Tensor(" << detach_var->Name()
                     << ") share data with " << self.Name();
             return detach_var;
           },
           py::return_value_policy::take_ownership, R"DOC(
1453

1454
        Returns a new Tensor, detached from the current graph.
Z
Zhou Wei 已提交
1455 1456
        It will share data with origin Tensor and always doesn't have a Tensor copy.
        In addition, the detached Tensor doesn't provide gradient propagation.
1457

1458
        Returns: The detached Tensor.
1459 1460 1461 1462

        Examples:
            .. code-block:: python

1463
                import paddle
Z
Zhou Wei 已提交
1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488

                x = paddle.to_tensor(1.0, stop_gradient=False)
                detach_x = x.detach()
                detach_x[:] = 10.0
                print(x)  # Tensor(shape=[1], dtype=float32, place=CPUPlace, stop_gradient=False,
                          #        [10.])
                y = x**2
                y.backward()
                print(x.grad)         # [20.0]
                print(detach_x.grad)  # None, 'stop_gradient=True' by default

                detach_x.stop_gradient = False # Set stop_gradient to be False, supported auto-grad
                z = detach_x**3
                z.backward()

                print(x.grad)         # [20.0], detach_x is detached from x's graph, not affect each other
                print(detach_x.grad)  # [300.0], detach_x has its own graph

                # Due to sharing of data with origin Tensor, There are some unsafe operations:
                y = 2 * x
                detach_x[:] = 5.0
                y.backward() 
                # It will raise Error:
                #   one of the variables needed for gradient computation has been modified by an inplace operation.
             
1489
       )DOC")
1490 1491
      .def("clear_gradient", &imperative::VarBase::ClearGradient,
           py::arg("set_to_zero") = true, R"DOC(
1492

1493
        Only for Tensor that has gradient, normally we use this for Parameters since other temporary Tensor doesen't has gradient.
1494

1495
        The Gradient of current Tensor will be set to ``0`` .
1496 1497 1498 1499 1500 1501

        Returns:  None

        Examples:
             .. code-block:: python

1502
                import paddle
Z
Zhou Wei 已提交
1503 1504 1505 1506 1507 1508 1509
                input = paddle.uniform([10, 2])
                linear = paddle.nn.Linear(2, 3)
                out = linear(input)
                out.backward()
                print("Before clear_gradient, linear.weight.grad: {}".format(linear.weight.grad))
                linear.weight.clear_gradient()
                print("After clear_gradient, linear.weight.grad: {}".format(linear.weight.grad))
1510
      )DOC")
1511 1512 1513
      .def("_gradient_set_empty", &imperative::VarBase::_GradientSetEmpty,
           py::arg("set_is_empty") = true)
      .def("_is_gradient_set_empty", &imperative::VarBase::_IsGradientSetEmpty)
Z
Zhou Wei 已提交
1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561
      .def("clone",
           [](std::shared_ptr<imperative::VarBase> &self) {
             const auto &tensor = self->Var().Get<framework::LoDTensor>();
             PADDLE_ENFORCE_EQ(
                 tensor.IsInitialized(), true,
                 platform::errors::InvalidArgument(
                     "%s has not been initialized", self->Name()));
             auto tracer = imperative::GetCurrentTracer();
             auto new_var = std::make_shared<imperative::VarBase>(
                 true, tracer->GenerateUniqueName(self->Name() + "_clone"));
             framework::AttributeMap attrs;
             imperative::NameVarBaseMap ins = {{"X", {self}}};
             imperative::NameVarBaseMap outs = {{"Out", {new_var}}};
             tracer->TraceOp("assign", ins, outs, attrs);
             return new_var;
           },
           py::return_value_policy::copy, R"DOC(

        Returns a new Tensor, which is clone of origin Tensor, and it remains in the current graph.
        It will always have a Tensor copy.
        Tn addition, the cloned Tensor provides gradient propagation.

        Returns: The cloned Tensor.

        Examples:
            .. code-block:: python

              import paddle

              x = paddle.to_tensor(1.0, stop_gradient=False)
              clone_x = x.clone()
              y = clone_x**2
              y.backward()
              print(clone_x.stop_gradient) # False
              print(clone_x.grad)          # [2.0], support gradient propagation
              print(x.stop_gradient)       # False
              print(x.grad)                # [2.0], clone_x support gradient propagation for x

              x = paddle.to_tensor(1.0)
              clone_x = x.clone()
              clone_x.stop_gradient = False
              z = clone_x**3
              z.backward()
              print(clone_x.stop_gradient) # False
              print(clone_x.grad)          # [3.0], support gradient propagation
              print(x.stop_gradient) # True
              print(x.grad)          # None
       )DOC")
L
Leo Chen 已提交
1562 1563 1564 1565 1566 1567
      .def("_grad_name", &imperative::VarBase::GradVarName)
      .def("_grad_value",
           [](imperative::VarBase &self) {
             return self.MutableGradVar()->Get<framework::LoDTensor>();
           },
           py::return_value_policy::reference)
1568 1569 1570 1571
      .def("_set_grad_type",
           [](imperative::VarBase &self, framework::proto::VarType::Type type) {
             self.MutableGradVarBase()->SetType(type);
           })
1572
      .def("_grad_ivar",
J
Jiabin Yang 已提交
1573 1574
           [](const imperative::VarBase &self) {
             auto &grad_var = self.GradVarBase();
1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585
             if (grad_var && grad_var->Var().IsInitialized()) {
               auto *tensor =
                   grad_var->MutableVar()->IsType<framework::LoDTensor>()
                       ? grad_var->MutableVar()
                             ->GetMutable<framework::LoDTensor>()
                       : grad_var->MutableVar()
                             ->GetMutable<framework::SelectedRows>()
                             ->mutable_value();
               if (tensor->IsInitialized()) {
                 return grad_var;
               }
J
Jiabin Yang 已提交
1586
             }
1587
             return std::shared_ptr<imperative::VarBase>(nullptr);
J
Jiabin Yang 已提交
1588 1589
           },
           py::return_value_policy::copy)
C
chentianyu03 已提交
1590 1591 1592 1593
      .def("_set_grad_ivar",
           [](imperative::VarBase &self, imperative::VarBase &grad) {
             self.SetGradVarBase(grad);
           })
1594 1595 1596 1597 1598 1599 1600 1601
      .def("_is_sparse",
           [](imperative::VarBase &self) {
             return self.Var().IsType<framework::SelectedRows>();
           })
      .def("_allreduce",
           [](imperative::VarBase &self,
              const imperative::ParallelStrategy &strategy) {
             if (strategy.nranks_ > 1) {
1602
#if defined(PADDLE_WITH_NCCL) || defined(PADDLE_WITH_RCCL)
1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619
#if NCCL_VERSION_CODE >= 2212
               imperative::AllReduce(self.Var(), self.MutableVar(), strategy);
#else
               if (!self.Var().IsType<framework::SelectedRows>()) {
                 imperative::AllReduce(self.Var(), self.MutableVar(), strategy);
               } else {
                 PADDLE_THROW(platform::errors::Unimplemented(
                     "Imperative SelectedRows allreduce is not supported when "
                     "paddle is compiled with NCCL verison lower than v2.2.12. "
                     "You can set is_sparse=False for the Layer containing "
                     "this argument, such as Embedding(is_sparse=False)."));
               }
#endif  // NCCL_VERSION_CODE
#else
               PADDLE_THROW(platform::errors::Unimplemented(
                   "Imperative allreduce is not supported when paddle is "
                   "not compiled with NCCL."));
1620
#endif  // PADDLE_WITH_NCCL or PADDLE_WITH_RCCL
1621 1622 1623
             }
           },
           py::call_guard<py::gil_scoped_release>())
1624 1625 1626
      .def("_register_grad_hook",
           [](imperative::VarBase &self, const py::handle &hook) {
             PADDLE_ENFORCE_EQ(
1627
                 !self.OverridedStopGradient() && self.HasGradVar(), true,
1628
                 platform::errors::InvalidArgument(
1629 1630 1631
                     "Cannot register gradient hook on a Tensor that stop "
                     "gradient or without gradient."));
             return self.GradVarBase()->AddVariableWrapperHook(
1632 1633 1634 1635 1636
                 std::make_shared<PyVariableWrapperHook>(hook.ptr()));
           })
      .def("_remove_grad_hook",
           [](imperative::VarBase &self, int64_t hook_id) {
             PADDLE_ENFORCE_EQ(
1637
                 !self.OverridedStopGradient() && self.HasGradVar(), true,
1638
                 platform::errors::InvalidArgument(
1639 1640 1641
                     "Cannot remove gradient hook on a Tensor that stop "
                     "gradient or without gradient."));
             return self.GradVarBase()->RemoveVariableWrapperHook(hook_id);
1642
           })
1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678
      .def("_register_backward_hook",
           [](imperative::VarBase &self, const py::handle &hook) {
             PADDLE_ENFORCE_EQ(
                 self.IsLeaf(), true,
                 platform::errors::InvalidArgument(
                     "Only can register backward hook for leaf Tensor."));
             PADDLE_ENFORCE_EQ(
                 !self.OverridedStopGradient() && self.HasGradVar(), true,
                 platform::errors::InvalidArgument(
                     "Cannot register backward hook on a Tensor that stop "
                     "gradient or without gradient."));
             auto py_func = PyObjectCast<std::function<void()>>(hook.ptr());
             self.GradVarBase()->AddVoidHook(
                 std::make_shared<std::function<void()>>(py_func));
           },
           R"DOC(
             Registers a backward hook for current Tensor.

             This hook will be called every time the gradient of current Tensor has been fully calculated.

             There are two differences with `_register_grad_hook`:
             1. This backward hook will be executed after the gradient accumulation completed across batchs,
                but the hook registered by `_register_grad_hook` will be executed the gradient accumulation
                completed in current batch.
             2. This backward hook function should have the following signature:

                  hook() -> None

                It requires no input and no return value.

             Args:
                 hook(function): A backward hook to be registered for Tensor.gradient

             Returns:
                 None
           )DOC")
1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706
      .def("cpu",
           [](const std::shared_ptr<imperative::VarBase> &self) {
             if (platform::is_cpu_place(self->Place())) {
               return self;
             } else {
               auto new_var = self->NewVarBase(platform::CPUPlace(), true);
               new_var->SetOverridedStopGradient(self->OverridedStopGradient());
               return new_var;
             }
           },
           R"DOC(
        Returns a copy of this Tensor in CPU memory.

        If this Tensor is already in CPU memory, then no copy is performed and the original Tensor is returned.

        Examples:
            .. code-block:: python

              import paddle
              x = paddle.to_tensor(1.0, place=paddle.CUDAPlace(0))
              print(x.place)    # CUDAPlace(0)
              
              y = x.cpu()
              print(y.place)    # CPUPlace

              )DOC")
      .def("pin_memory",
           [](const std::shared_ptr<imperative::VarBase> &self) {
1707
#if !defined(PADDLE_WITH_CUDA) && !defined(PADDLE_WITH_HIP)
1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738
             PADDLE_THROW(platform::errors::PermissionDenied(
                 "Cannot copy this Tensor to pinned memory in CPU version "
                 "Paddle, "
                 "Please recompile or reinstall Paddle with CUDA support."));
#endif
             if (platform::is_cuda_pinned_place(self->Place())) {
               return self;
             } else {
               auto new_var =
                   self->NewVarBase(platform::CUDAPinnedPlace(), true);
               new_var->SetOverridedStopGradient(self->OverridedStopGradient());
               return new_var;
             }
           },
           R"DOC(
        Returns a copy of this Tensor in pin memory.

        If this Tensor is already in pin memory, then no copy is performed and the original Tensor is returned.

        Examples:
            .. code-block:: python

              import paddle
              x = paddle.to_tensor(1.0, place=paddle.CUDAPlace(0))
              print(x.place)      # CUDAPlace(0)

              y = x.pin_memory()
              print(y.place)      # CUDAPinnedPlace

      )DOC")
      .def("cuda",
1739 1740
           [](const std::shared_ptr<imperative::VarBase> &self,
              py::handle &handle, bool blocking) {
1741
#if !defined(PADDLE_WITH_CUDA) && !defined(PADDLE_WITH_HIP)
1742 1743 1744 1745 1746
             PADDLE_THROW(platform::errors::PermissionDenied(
                 "Cannot copy this Tensor to GPU in CPU version Paddle, "
                 "Please recompile or reinstall Paddle with CUDA support."));
#else
             int device_count = platform::GetCUDADeviceCount();
1747 1748
             int device_id = 0;
             if (handle == py::none()) {
1749 1750 1751
               if (platform::is_gpu_place(self->Place())) {
                 return self;
               }
1752 1753 1754 1755 1756 1757 1758
             } else {
               PyObject *py_obj = handle.ptr();
               PADDLE_ENFORCE_EQ(
                   PyCheckInteger(py_obj), true,
                   platform::errors::InvalidArgument(
                       " 'device_id' must be a positive integer"));
               device_id = py::cast<int>(handle);
1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781
             }
             PADDLE_ENFORCE_GE(
                 device_id, 0,
                 platform::errors::InvalidArgument(
                     "Can not copy Tensor to Invalid CUDAPlace(%d), device id "
                     "must inside [0, %d)",
                     device_id, device_count));
             PADDLE_ENFORCE_LT(
                 device_id, device_count,
                 platform::errors::InvalidArgument(
                     "Can not copy Tensor to Invalid CUDAPlace(%d), device id "
                     "must inside [0, %d)",
                     device_id, device_count));
             platform::CUDAPlace place = platform::CUDAPlace(device_id);
             if (platform::is_same_place(self->Place(), place)) {
               return self;
             } else {
               auto new_var = self->NewVarBase(place, blocking);
               new_var->SetOverridedStopGradient(self->OverridedStopGradient());
               return new_var;
             }
#endif
           },
1782
           py::arg("device_id") = py::none(), py::arg("blocking") = true, R"DOC(
1783 1784 1785 1786 1787 1788
        Returns a copy of this Tensor in GPU memory.

        If this Tensor is already in GPU memory and device_id is default, 
        then no copy is performed and the original Tensor is returned.
        
        Args:
1789
            device_id(int, optional): The destination GPU device id. Default: None, means current device.
1790 1791 1792 1793 1794 1795
            blocking(bool, optional): If False and the source is in pinned memory, the copy will be 
              asynchronous with respect to the host. Otherwise, the argument has no effect. Default: False.

        Examples:
            .. code-block:: python

1796
              # required: gpu
1797 1798 1799 1800 1801 1802
              import paddle
              x = paddle.to_tensor(1.0, place=paddle.CPUPlace())
              print(x.place)        # CPUPlace

              y = x.cuda()
              print(y.place)        # CUDAPlace(0)
1803 1804 1805
            
              y = x.cuda(None)
              print(y.place)        # CUDAPlace(0)
1806 1807 1808 1809

              y = x.cuda(1)
              print(y.place)        # CUDAPlace(1)
       )DOC")
K
Kaipeng Deng 已提交
1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838
      .def("_share_memory",
           [](const std::shared_ptr<imperative::VarBase> &self) {
#ifndef _WIN32
             PADDLE_ENFORCE_EQ(
                 platform::is_cpu_place(self->Place()), true,
                 platform::errors::InvalidArgument(
                     "Sharing memory only support CPU Tensor currently"));
             // 1. get LoDTensor
             auto *t = self->MutableVar()->GetMutable<framework::LoDTensor>();
             // 2. allocate shared memory
             void *data_ptr = t->data<void>();
             size_t data_size = t->numel() * framework::SizeOfType(t->type());
             auto shared_writer_holder =
                 memory::allocation::AllocateMemoryMapWriterAllocation(
                     data_size);
             // 3. maintain mmap fd set & backup ipc_name
             const std::string &ipc_name = shared_writer_holder->ipc_name();
             memory::allocation::MemoryMapFdSet::Instance().Insert(ipc_name);
             // 4. copy data & reset holder
             memory::Copy(platform::CPUPlace(), shared_writer_holder->ptr(),
                          platform::CPUPlace(), data_ptr, data_size);
             t->ResetHolder(shared_writer_holder);
             return *t;
#else
             PADDLE_THROW(platform::errors::PermissionDenied(
                 "Sharing memory in Windows OS is not supported currently"));
#endif
           },
           py::return_value_policy::reference)
1839
      .def("copy_", &imperative::VarBase::CopyFrom)
1840
      .def("_copy_to",
1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856
           [](const std::shared_ptr<imperative::VarBase> &self,
              const platform::CPUPlace &place, bool blocking) {
             auto new_var = self->NewVarBase(place, blocking);
             // Note(zhiqiu): Since NewVarBase may use GpuCopyAsync to
             // copy data from the tensor of self to the tensor of new varbase,
             // we need to ensure that the varbase self is not destructed until
             // the GpuCopyAsync is completed. Otherwise, the memory may be
             // freed
             // when varbase self is destructed.
             // To do that, we increase the reference count of self by 1 and
             // add a cuda event to wait the GpuCopyAsync's completion.
             if (!blocking) {
               IncreaseVarbaseReferenceCountUntilCopyComplete(self, place);
             }
             return new_var;
           },
J
Jiabin Yang 已提交
1857
           py::return_value_policy::copy)
1858
      .def("_copy_to",
1859 1860 1861 1862 1863 1864 1865 1866
           [](const std::shared_ptr<imperative::VarBase> &self,
              const platform::CUDAPinnedPlace &place, bool blocking) {
             auto new_var = self->NewVarBase(place, blocking);
             if (!blocking) {
               IncreaseVarbaseReferenceCountUntilCopyComplete(self, place);
             }
             return new_var;
           },
1867
           py::return_value_policy::copy)
1868
      .def("_copy_to",
1869 1870 1871 1872 1873 1874 1875 1876
           [](const std::shared_ptr<imperative::VarBase> &self,
              const platform::XPUPlace &place, bool blocking) {
             auto new_var = self->NewVarBase(place, blocking);
             if (!blocking) {
               IncreaseVarbaseReferenceCountUntilCopyComplete(self, place);
             }
             return new_var;
           },
1877
           py::return_value_policy::copy)
1878
      .def("_copy_to",
1879 1880 1881 1882 1883 1884 1885 1886
           [](const std::shared_ptr<imperative::VarBase> &self,
              const platform::CUDAPlace &place, bool blocking) {
             auto new_var = self->NewVarBase(place, blocking);
             if (!blocking) {
               IncreaseVarbaseReferenceCountUntilCopyComplete(self, place);
             }
             return new_var;
           },
J
Jiabin Yang 已提交
1887
           py::return_value_policy::copy)
1888 1889 1890 1891 1892 1893 1894 1895 1896 1897
      .def("_copy_to",
           [](const std::shared_ptr<imperative::VarBase> &self,
              const platform::NPUPlace &place, bool blocking) {
             auto new_var = self->NewVarBase(place, blocking);
             if (!blocking) {
               IncreaseVarbaseReferenceCountUntilCopyComplete(self, place);
             }
             return new_var;
           },
           py::return_value_policy::copy)
C
chentianyu03 已提交
1898 1899 1900 1901 1902 1903 1904 1905 1906 1907
      .def("_copy_to",
           [](const std::shared_ptr<imperative::VarBase> &self,
              const platform::Place &place, bool blocking) {
             auto new_var = self->NewVarBase(place, blocking);
             if (!blocking) {
               IncreaseVarbaseReferenceCountUntilCopyComplete(self, place);
             }
             return new_var;
           },
           py::return_value_policy::copy)
J
Jiabin Yang 已提交
1908
      .def("value", [](imperative::VarBase &self) { return self.MutableVar(); },
1909
           py::return_value_policy::reference)
1910 1911 1912
      .def("_clear",
           [](const std::shared_ptr<imperative::VarBase> &self) {
             auto *t = self->MutableVar()->GetMutable<framework::LoDTensor>();
1913 1914 1915 1916
             PADDLE_ENFORCE_EQ(
                 t->IsInitialized(), true,
                 platform::errors::InvalidArgument(
                     "Tensor %s has not been initialized!", self->Name()));
1917 1918 1919 1920 1921
             t->clear();
           })
      .def("_offset",
           [](const std::shared_ptr<imperative::VarBase> &self) {
             auto *t = self->MutableVar()->GetMutable<framework::LoDTensor>();
1922 1923 1924 1925
             PADDLE_ENFORCE_EQ(
                 t->IsInitialized(), true,
                 platform::errors::InvalidArgument(
                     "Tensor %s has not been initialized!", self->Name()));
1926 1927
             return t->offset();
           })
1928
      .def("_share_buffer_to",
1929
           [](const std::shared_ptr<imperative::VarBase> &self,
1930 1931 1932 1933 1934 1935 1936 1937
              std::shared_ptr<imperative::VarBase> &dst) {
             auto *src = self->MutableVar()->GetMutable<framework::LoDTensor>();
             auto *dst_ = dst->MutableVar()->GetMutable<framework::LoDTensor>();
             PADDLE_ENFORCE_EQ(
                 src->IsInitialized(), true,
                 platform::errors::InvalidArgument(
                     "Tensor %s has not been initialized!", self->Name()));
             dst_->ShareBufferWith(*src);
1938 1939 1940
           })
      .def("_is_shared_buffer_with",
           [](const std::shared_ptr<imperative::VarBase> &self,
1941 1942 1943 1944 1945 1946 1947
              std::shared_ptr<imperative::VarBase> &dst) {
             auto *src = self->MutableVar()->GetMutable<framework::LoDTensor>();
             auto *dst_ = dst->MutableVar()->GetMutable<framework::LoDTensor>();
             if (!src->IsInitialized() || !dst_->IsInitialized()) {
               return false;
             }
             return dst_->IsSharedBufferWith(*src);
1948 1949 1950 1951 1952
           })
      .def("_slice",
           [](const std::shared_ptr<imperative::VarBase> &self,
              int64_t begin_idx, int64_t end_idx) {
             auto *t = self->MutableVar()->GetMutable<framework::LoDTensor>();
1953 1954 1955 1956
             PADDLE_ENFORCE_EQ(
                 t->IsInitialized(), true,
                 platform::errors::InvalidArgument(
                     "Tensor %s has not been initialized!", self->Name()));
1957 1958 1959 1960 1961 1962 1963 1964
             return t->Slice(begin_idx, end_idx);
           })
      .def("_copy_gradient_from",
           [](std::shared_ptr<imperative::VarBase> &self,
              const imperative::VarBase &src) { self->_CopyGradientFrom(src); })
      .def("_numel",
           [](std::shared_ptr<imperative::VarBase> &self) {
             auto *t = self->MutableVar()->GetMutable<framework::LoDTensor>();
1965 1966 1967 1968
             PADDLE_ENFORCE_EQ(
                 t->IsInitialized(), true,
                 platform::errors::InvalidArgument(
                     "Tensor %s has not been initialized!", self->Name()));
1969 1970
             return t->numel();
           })
1971 1972
      .def_property("name", &imperative::VarBase::Name,
                    &imperative::VarBase::SetName)
L
Leo Chen 已提交
1973 1974 1975 1976 1977
      .def_property("stop_gradient",
                    &imperative::VarBase::OverridedStopGradient,
                    &imperative::VarBase::SetOverridedStopGradient)
      .def_property("persistable", &imperative::VarBase::Persistable,
                    &imperative::VarBase::SetPersistable)
1978 1979 1980 1981 1982 1983 1984 1985 1986
      .def_property_readonly(
          "shape",
          [](imperative::VarBase &self) {
            if (self.Var().IsType<framework::LoDTensor>()) {
              return framework::vectorize<int>(
                  self.Var().Get<framework::LoDTensor>().dims());
            } else if (self.Var().IsType<framework::SelectedRows>()) {
              return framework::vectorize<int>(
                  self.Var().Get<framework::SelectedRows>().value().dims());
S
Steffy-zxf 已提交
1987 1988 1989 1990 1991 1992
            } else if (self.Var().IsType<framework::Strings>()) {
              return std::vector<int>{static_cast<int>(
                  self.Var().Get<framework::Strings>().size())};
            } else if (self.Var().IsType<framework::Vocab>()) {
              return std::vector<int>{
                  static_cast<int>(self.Var().Get<framework::Vocab>().size())};
1993 1994 1995 1996 1997 1998 1999
            } else {
              VLOG(2) << "It is meaningless to get shape of "
                         "variable type "
                      << GetTypeName(self);
              return std::vector<int>();
            }
          })
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028
      .def_property_readonly("is_leaf", &imperative::VarBase::IsLeaf,
                             R"DOC(
      Whether a Tensor is leaf Tensor.

      For the Tensor whose stop_gradient is ``True`` , it will be leaf Tensor. 
      
      For the Tensor whose stop_gradient is ``False`` , it will be leaf Tensor too if it is created by user.

      Returns:
          bool: Whether a Tensor is leaf Tensor.

      Examples:
          .. code-block:: python

              import paddle

              x = paddle.to_tensor(1.)
              print(x.is_leaf) # True

              x = paddle.to_tensor(1., stop_gradient=True)
              y = x + 1
              print(x.is_leaf) # True
              print(y.is_leaf) # True

              x = paddle.to_tensor(1., stop_gradient=False)
              y = x + 1
              print(x.is_leaf) # True
              print(y.is_leaf) # False
       )DOC")
2029 2030 2031
      .def_property_readonly(
          "place", [](imperative::VarBase &self) { return self.Place(); },
          py::return_value_policy::copy)
2032 2033 2034 2035 2036 2037
      .def_property_readonly("_place_str",
                             [](imperative::VarBase &self) {
                               std::stringstream ostr;
                               ostr << self.Place();
                               return ostr.str();
                             })
J
Jiabin Yang 已提交
2038
      .def_property_readonly("type", &imperative::VarBase::Type)
L
Leo Chen 已提交
2039
      .def_property_readonly("dtype", &imperative::VarBase::DataType);
2040

2041 2042 2043 2044 2045
  // NOTE(zhiqiu): set the metaclass of Layer.
  // See details: https://github.com/pybind/pybind11/pull/679
  // https://github.com/pybind/pybind11/blob/028812ae7eee307dca5f8f69d467af7b92cc41c8/tests/test_methods_and_attributes.cpp#L284
  py::class_<imperative::Layer, Layer /* <--- trampoline*/> layer(
      m, "Layer", py::metaclass((PyObject *)&PyType_Type));  // NOLINT
2046
  layer.def(py::init<>())
2047 2048 2049 2050 2051
      .def("forward",
           [](imperative::Layer &self,
              const std::vector<std::shared_ptr<imperative::VarBase>> &inputs) {
             return self.Forward(inputs);
           });
2052

2053 2054 2055 2056 2057
  py::class_<imperative::jit::ProgramDescTracer>(m, "ProgramDescTracer", "")
      .def("create_program_desc",
           &imperative::jit::ProgramDescTracer::CreateProgramDesc)
      .def("reset", &imperative::jit::ProgramDescTracer::Reset);

L
Leo Chen 已提交
2058 2059 2060 2061 2062 2063 2064
  py::enum_<paddle::imperative::AmpLevel>(m, "AmpLevel", py::arithmetic())
      .value("O0", paddle::imperative::AmpLevel::O0)
      .value("O1", paddle::imperative::AmpLevel::O1)
      .value("O2", paddle::imperative::AmpLevel::O2)
      .value("O3", paddle::imperative::AmpLevel::O3)
      .export_values();

2065
  py::class_<imperative::Tracer, std::shared_ptr<imperative::Tracer>>(
2066
      m, "Tracer", R"DOC()DOC")
2067
      .def("__init__",
J
Jiabin Yang 已提交
2068
           [](imperative::Tracer &self) { new (&self) imperative::Tracer(); })
2069 2070 2071
      .def_property("_enable_program_desc_tracing",
                    &imperative::Tracer::IsProgramDescTracingEnabled,
                    &imperative::Tracer::SetEnableProgramDescTracing)
L
Leo Chen 已提交
2072 2073
      .def_property("_amp_level", &imperative::Tracer::GetAmpLevel,
                    &imperative::Tracer::SetAmpLevel)
2074
      .def_property("_has_grad", &imperative::Tracer::HasGrad,
2075
                    &imperative::Tracer::SetHasGrad)
2076 2077 2078 2079 2080 2081 2082 2083
      .def_property(
          "_expected_place",
          [](const imperative::Tracer &self) -> py::object {
            return py::cast(self.ExpectedPlace());
          },
          [](imperative::Tracer &self, const py::object &obj) {
            if (py::isinstance<platform::CUDAPlace>(obj)) {
              auto p = obj.cast<platform::CUDAPlace *>();
L
Leo Chen 已提交
2084
              self.SetExpectedPlace(*p);
2085 2086
              VLOG(4) << "Tracer(" << &self << ")"
                      << " set expected place " << *p;
2087 2088 2089
            } else if (py::isinstance<platform::XPUPlace>(obj)) {
              auto p = obj.cast<platform::XPUPlace *>();
              self.SetExpectedPlace(*p);
2090 2091
              VLOG(4) << "Tracer(" << &self << ")"
                      << " set expected place " << *p;
2092 2093
            } else if (py::isinstance<platform::CPUPlace>(obj)) {
              auto p = obj.cast<platform::CPUPlace *>();
L
Leo Chen 已提交
2094
              self.SetExpectedPlace(*p);
2095 2096
              VLOG(4) << "Tracer(" << &self << ")"
                      << " set expected place " << *p;
2097 2098
            } else if (py::isinstance<platform::CUDAPinnedPlace>(obj)) {
              auto p = obj.cast<platform::CUDAPinnedPlace *>();
L
Leo Chen 已提交
2099
              self.SetExpectedPlace(*p);
2100 2101
              VLOG(4) << "Tracer(" << &self << ")"
                      << " set expected place " << *p;
2102 2103 2104 2105 2106
            } else if (py::isinstance<platform::NPUPlace>(obj)) {
              auto p = obj.cast<platform::NPUPlace *>();
              self.SetExpectedPlace(*p);
              VLOG(4) << "Tracer(" << &self << ")"
                      << " set expected place " << *p;
2107 2108 2109 2110 2111
            } else if (py::isinstance<platform::Place>(obj)) {
              auto p = obj.cast<platform::Place *>();
              self.SetExpectedPlace(*p);
              VLOG(4) << "Tracer(" << &self << ")"
                      << " set expected place " << *p;
2112
            } else {
L
Leo Chen 已提交
2113
              PADDLE_THROW(platform::errors::InvalidArgument(
2114
                  "Incompatible Place Type: supports XPUPlace, CUDAPlace, "
2115
                  "CPUPlace, NPUPlace"
L
Leo Chen 已提交
2116 2117
                  "and CUDAPinnedPlace, "
                  "but got Unknown Type!"));
2118 2119
            }
          })
2120 2121 2122
      .def("_get_program_desc_tracer",
           &imperative::Tracer::GetProgramDescTracer,
           py::return_value_policy::reference)
2123
      .def("_generate_unique_name", &imperative::Tracer::GenerateUniqueName,
2124
           py::arg("key") = "dygraph_tmp")
2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137 2138 2139 2140
      .def("_set_amp_op_list",
           [](imperative::Tracer &self,
              std::unordered_set<std::string> &allow_ops,
              std::unordered_set<std::string> &block_ops) {
             // NOTE(zhiqiu): The automatic conversion in pybind11 between
             // c++
             // STL and python set/list/dict involve a copy operation that
             // prevents pass-by-reference semantics, so it is ok to swap.
             // The reaseon why not directly pass
             // std::shared_ptr<std::unordered_set<std::string>>
             // is that pybind11 forbid shared_ptr<T> where T is not custom
             // type.
             imperative::AmpOperators::Instance().GetMutableAllowOps()->swap(
                 allow_ops);
             imperative::AmpOperators::Instance().GetMutableBlockOps()->swap(
                 block_ops);
2141
             VLOG(5) << "AMP operators changed, "
2142 2143
                     << imperative::AmpOperators::Instance();
           })
2144 2145 2146
      .def("_get_amp_op_list",
           [](imperative::Tracer &self) {
             return std::make_tuple(
2147 2148
                 *(imperative::AmpOperators::Instance().GetMutableAllowOps()),
                 *(imperative::AmpOperators::Instance().GetMutableBlockOps()));
2149
           })
2150 2151 2152 2153 2154 2155 2156 2157 2158 2159 2160 2161 2162
      .def("trace",
           [](imperative::Tracer &self, const std::string &type,
              const PyNameVarBaseMap &ins, const PyNameVarBaseMap &outs,
              framework::AttributeMap attrs, const platform::XPUPlace &place,
              bool trace_backward) {
             auto ins_map = ConvertToNameVarBaseMap(ins);
             auto outs_map = ConvertToNameVarBaseMap(outs);
             {
               py::gil_scoped_release release;
               self.TraceOp(type, std::move(ins_map), std::move(outs_map),
                            std::move(attrs), place, trace_backward);
             }
           })
M
minqiyang 已提交
2163
      .def("trace",
J
Jiabin Yang 已提交
2164 2165 2166 2167 2168 2169
           [](imperative::Tracer &self, const std::string &type,
              const PyNameVarBaseMap &ins, const PyNameVarBaseMap &outs,
              framework::AttributeMap attrs, const platform::CUDAPlace &place,
              bool trace_backward) {
             auto ins_map = ConvertToNameVarBaseMap(ins);
             auto outs_map = ConvertToNameVarBaseMap(outs);
2170 2171
             {
               py::gil_scoped_release release;
J
Jiabin Yang 已提交
2172 2173
               self.TraceOp(type, std::move(ins_map), std::move(outs_map),
                            std::move(attrs), place, trace_backward);
2174
             }
M
minqiyang 已提交
2175
           })
2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188
      .def("trace",
           [](imperative::Tracer &self, const std::string &type,
              const PyNameVarBaseMap &ins, const PyNameVarBaseMap &outs,
              framework::AttributeMap attrs, const platform::NPUPlace &place,
              bool trace_backward) {
             auto ins_map = ConvertToNameVarBaseMap(ins);
             auto outs_map = ConvertToNameVarBaseMap(outs);
             {
               py::gil_scoped_release release;
               self.TraceOp(type, std::move(ins_map), std::move(outs_map),
                            std::move(attrs), place, trace_backward);
             }
           })
J
Jiabin Yang 已提交
2189 2190 2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201
      .def("trace",
           [](imperative::Tracer &self, const std::string &type,
              const PyNameVarBaseMap &ins, const PyNameVarBaseMap &outs,
              framework::AttributeMap attrs, const platform::CPUPlace &place,
              bool trace_backward) {
             auto ins_map = ConvertToNameVarBaseMap(ins);
             auto outs_map = ConvertToNameVarBaseMap(outs);
             {
               py::gil_scoped_release release;
               self.TraceOp(type, std::move(ins_map), std::move(outs_map),
                            std::move(attrs), place, trace_backward);
             }
           });
2202 2203

  // define parallel context
2204 2205 2206
  py::class_<imperative::ParallelStrategy> parallel_strategy(
      m, "ParallelStrategy", "");
  parallel_strategy.def(py::init())
2207 2208
      .def_property(
          "nranks",
2209 2210
          [](const imperative::ParallelStrategy &self) { return self.nranks_; },
          [](imperative::ParallelStrategy &self, int nranks) {
2211 2212 2213
            self.nranks_ = nranks;
          })
      .def_property("local_rank",
2214
                    [](const imperative::ParallelStrategy &self) {
2215 2216
                      return self.local_rank_;
                    },
2217
                    [](imperative::ParallelStrategy &self, int local_rank) {
2218 2219 2220 2221
                      self.local_rank_ = local_rank;
                    })
      .def_property(
          "trainer_endpoints",
2222
          [](const imperative::ParallelStrategy &self) {
2223 2224
            return self.trainer_endpoints_;
          },
2225
          [](imperative::ParallelStrategy &self, std::vector<std::string> eps) {
2226 2227 2228
            self.trainer_endpoints_ = eps;
          })
      .def_property("current_endpoint",
2229
                    [](const imperative::ParallelStrategy &self) {
2230 2231
                      return self.current_endpoint_;
                    },
2232
                    [](imperative::ParallelStrategy &self,
2233 2234 2235 2236 2237 2238 2239
                       const std::string &ep) { self.current_endpoint_ = ep; })
      .def_property(
          "nrings",
          [](const imperative::ParallelStrategy &self) { return self.nrings_; },
          [](imperative::ParallelStrategy &self, int nrings) {
            self.nrings_ = nrings;
          });
2240

2241 2242 2243 2244
  m.def("varbase_copy", &VarBaseCopy<platform::Place>);
  m.def("varbase_copy", &VarBaseCopy<platform::CPUPlace>);
  m.def("varbase_copy", &VarBaseCopy<platform::CUDAPlace>);
  m.def("varbase_copy", &VarBaseCopy<platform::XPUPlace>);
2245
  m.def("varbase_copy", &VarBaseCopy<platform::CUDAPinnedPlace>);
2246
  m.def("varbase_copy", &VarBaseCopy<platform::NPUPlace>);
2247

2248 2249 2250 2251 2252 2253 2254
  m.def(
      "dygraph_partial_grad",
      [](const std::vector<std::shared_ptr<imperative::VarBase>> &input_targets,
         const std::vector<std::shared_ptr<imperative::VarBase>>
             &output_targets,
         const std::vector<std::shared_ptr<imperative::VarBase>> &output_grads,
         const std::vector<std::shared_ptr<imperative::VarBase>> &no_grad_vars,
2255 2256
         const platform::Place &place, bool create_graph, bool retain_graph,
         bool allow_unused, bool only_inputs) {
Z
Zeng Jinle 已提交
2257 2258
        imperative::PartialGradEngine engine(
            input_targets, output_targets, output_grads, no_grad_vars, place,
2259
            create_graph, retain_graph, allow_unused, only_inputs);
2260 2261 2262 2263 2264
        engine.Execute();
        return engine.GetResult();
      },
      py::call_guard<py::gil_scoped_release>());

2265 2266 2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277
  m.def(
      "dygraph_run_backward",
      [](const std::vector<std::shared_ptr<imperative::VarBase>> &tensors,
         const std::vector<std::shared_ptr<imperative::VarBase>> &grad_tensors,
         bool retain_graph, const imperative::Tracer &tracer) {
        auto *engine = tracer.GetEngine();
        engine->Init(tensors, grad_tensors, retain_graph);
        VLOG(3) << "Start backward";
        engine->Execute();
        VLOG(3) << "Finish backward";
      },
      py::call_guard<py::gil_scoped_release>());

2278
#if defined(PADDLE_WITH_NCCL) || defined(PADDLE_WITH_RCCL) || \
2279
    defined(PADDLE_WITH_XPU_BKCL) || defined(PADDLE_WITH_GLOO)
2280 2281 2282 2283 2284 2285
  py::class_<imperative::ParallelContext,
             std::shared_ptr<imperative::ParallelContext>>(m,
                                                           "ParallelContext");

  py::class_<imperative::Reducer, std::shared_ptr<imperative::Reducer>>(
      m, "Reducer", R"DOC()DOC")
S
ShenLiang 已提交
2286 2287 2288 2289 2290
      .def(py::init<const std::vector<std::shared_ptr<imperative::VarBase>> &,
                    const std::vector<std::vector<size_t>> &,
                    const std::vector<bool> &,
                    std::shared_ptr<imperative::ParallelContext>,
                    const std::vector<size_t> &, bool>())
2291
      .def("prepare_for_backward", &imperative::Reducer::PrepareForBackward,
2292
           py::arg("vars"), py::call_guard<py::gil_scoped_release>());
2293 2294 2295 2296

  m.def("assign_group_by_size", &imperative::AssignGroupBySize, py::arg("vars"),
        py::arg("is_sparse_gradient"),
        py::arg("group_size_limits") = std::vector<size_t>{25 * 1024 * 1024},
2297
        py::arg("tensor_indices") = std::vector<int64_t>{},
2298
        py::call_guard<py::gil_scoped_release>());
2299
#endif
2300

2301
#if defined(PADDLE_WITH_NCCL) || defined(PADDLE_WITH_RCCL)
2302 2303 2304 2305 2306
  py::class_<imperative::NCCLParallelContext, imperative::ParallelContext,
             std::shared_ptr<imperative::NCCLParallelContext>>(
      m, "NCCLParallelContext")
      .def(py::init<const imperative::ParallelStrategy &,
                    const platform::CUDAPlace &>())
K
kuizhiqing 已提交
2307 2308 2309 2310
      .def("init", [](imperative::NCCLParallelContext &self) { self.Init(); })
      .def("init_with_ring_id",
           &imperative::NCCLParallelContext::InitWithRingID,
           py::arg("ring_id"));
2311 2312 2313 2314 2315 2316 2317 2318
#endif

#if defined(PADDLE_WITH_XPU_BKCL)
  py::class_<imperative::BKCLParallelContext, imperative::ParallelContext,
             std::shared_ptr<imperative::BKCLParallelContext>>(
      m, "BKCLParallelContext")
      .def(py::init<const imperative::ParallelStrategy &,
                    const platform::XPUPlace &>())
K
kuizhiqing 已提交
2319 2320 2321 2322
      .def("init", [](imperative::BKCLParallelContext &self) { self.Init(); })
      .def("init_with_ring_id",
           &imperative::BKCLParallelContext::InitWithRingID,
           py::arg("ring_id"));
2323
#endif
2324 2325 2326 2327 2328 2329 2330 2331 2332 2333 2334 2335 2336 2337

#if defined(PADDLE_WITH_GLOO)
  // xiongkun
  py::class_<imperative::GLOOParallelContext, imperative::ParallelContext,
             std::shared_ptr<imperative::GLOOParallelContext>>(
      m, "GLOOParallelContext")
      .def(py::init<const imperative::ParallelStrategy &,
                    const platform::CPUPlace &>())
      .def("init", [](imperative::GLOOParallelContext &self) { self.Init(); })
      .def("init_with_ring_id",
           &imperative::GLOOParallelContext::InitWithRingID,
           py::arg("ring_id"));
#endif

2338 2339 2340 2341 2342 2343 2344 2345 2346 2347 2348 2349 2350 2351 2352 2353 2354 2355 2356 2357 2358 2359 2360
  m.def("pylayer_apply",
        [](const platform::CPUPlace &place, const py::object &cls,
           const py::args args, const py::kwargs kwargs) {
          return imperative::PyLayerApply(place, cls, args, kwargs);
        });

  m.def("pylayer_apply",
        [](const platform::CUDAPlace &place, const py::object &cls,
           const py::args args, const py::kwargs kwargs) {
          return imperative::PyLayerApply(place, cls, args, kwargs);
        });

  m.def("pylayer_apply",
        [](const platform::XPUPlace &place, const py::object &cls,
           const py::args args, const py::kwargs kwargs) {
          return imperative::PyLayerApply(place, cls, args, kwargs);
        });

  m.def("pylayer_apply",
        [](const platform::CUDAPinnedPlace &place, const py::object &cls,
           const py::args args, const py::kwargs kwargs) {
          return imperative::PyLayerApply(place, cls, args, kwargs);
        });
2361 2362 2363 2364 2365 2366

  m.def("pylayer_apply",
        [](const platform::NPUPlace &place, const py::object &cls,
           const py::args args, const py::kwargs kwargs) {
          return imperative::PyLayerApply(place, cls, args, kwargs);
        });
2367 2368 2369 2370 2371 2372 2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390 2391 2392 2393 2394 2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416 2417 2418 2419 2420 2421 2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437 2438 2439 2440 2441 2442 2443 2444 2445 2446 2447 2448 2449 2450 2451 2452 2453 2454 2455 2456 2457 2458 2459 2460 2461 2462 2463 2464 2465 2466 2467 2468 2469 2470 2471 2472 2473 2474 2475 2476 2477 2478 2479 2480 2481 2482 2483 2484 2485 2486 2487 2488 2489 2490 2491 2492 2493 2494 2495 2496 2497 2498 2499 2500 2501 2502 2503 2504 2505 2506 2507 2508 2509 2510 2511 2512 2513 2514 2515 2516 2517 2518 2519 2520 2521 2522 2523 2524 2525 2526 2527 2528 2529 2530 2531 2532 2533 2534 2535 2536 2537 2538 2539 2540 2541 2542 2543 2544 2545 2546 2547 2548 2549 2550 2551 2552 2553 2554 2555 2556 2557 2558 2559 2560 2561 2562 2563 2564 2565 2566 2567 2568 2569 2570 2571 2572 2573 2574 2575 2576 2577 2578 2579 2580 2581 2582 2583 2584 2585 2586 2587 2588 2589 2590 2591 2592 2593 2594 2595 2596 2597 2598 2599 2600 2601 2602 2603 2604 2605 2606 2607 2608 2609 2610 2611 2612 2613 2614 2615 2616 2617 2618 2619 2620 2621 2622 2623 2624 2625 2626 2627 2628 2629 2630 2631 2632 2633 2634 2635 2636 2637 2638 2639 2640 2641 2642 2643 2644 2645 2646 2647 2648 2649 2650 2651 2652 2653 2654 2655 2656 2657 2658 2659 2660 2661 2662 2663 2664 2665 2666 2667 2668 2669 2670 2671 2672 2673 2674 2675 2676 2677 2678 2679 2680 2681 2682 2683 2684 2685 2686 2687 2688 2689 2690 2691 2692 2693 2694 2695 2696 2697 2698 2699 2700 2701 2702 2703

#if defined(PADDLE_WITH_CUDA)
  m.def(
      "async_write",
      [](const imperative::VarBase &src, imperative::VarBase &dst,
         const imperative::VarBase &offset, const imperative::VarBase &count) {
        PADDLE_ENFORCE_EQ(
            platform::is_gpu_place(src.Place()), true,
            platform::errors::InvalidArgument(
                "Required `src` device should be CUDAPlace, but received %d. ",
                src.Place()));
        PADDLE_ENFORCE_EQ(
            platform::is_cuda_pinned_place(dst.Place()), true,
            platform::errors::InvalidArgument(
                "Required `dst` device should be CUDAPinnedPlace, "
                "but received %d. ",
                dst.Place()));
        PADDLE_ENFORCE_EQ(
            platform::is_cpu_place(offset.Place()), true,
            platform::errors::InvalidArgument("Required `offset` device should "
                                              "be CPUPlace, but received %d. ",
                                              offset.Place()));
        PADDLE_ENFORCE_EQ(
            platform::is_cpu_place(count.Place()), true,
            platform::errors::InvalidArgument(
                "Required `count` device should be CPUPlace, but received %d. ",
                count.Place()));

        // TODO(daisiming): In future, add index as arguments following
        // async_read.
        auto &src_tensor = src.Var().Get<framework::LoDTensor>();
        auto *dst_tensor = dst.MutableVar()->GetMutable<framework::LoDTensor>();
        auto &offset_tensor = offset.Var().Get<framework::LoDTensor>();
        auto &count_tensor = count.Var().Get<framework::LoDTensor>();
        const auto &deviceId = paddle::platform::GetCurrentDeviceId();

        PADDLE_ENFORCE_EQ(offset_tensor.dims().size(), 1,
                          platform::errors::InvalidArgument(
                              "`offset` tensor should be one-dimensional."));
        PADDLE_ENFORCE_EQ(count_tensor.dims().size(), 1,
                          platform::errors::InvalidArgument(
                              "`count` tensor should be one-dimensional."));
        PADDLE_ENFORCE_EQ(offset_tensor.numel(), count_tensor.numel(),
                          platform::errors::InvalidArgument(
                              "`offset` and `count` tensor size dismatch."));
        PADDLE_ENFORCE_EQ(
            src_tensor.dims().size(), dst_tensor->dims().size(),
            platform::errors::InvalidArgument(
                "`src` and `dst` should have the same tensor shape, "
                "except for the first dimension."));
        for (int i = 1; i < src_tensor.dims().size(); i++) {
          PADDLE_ENFORCE_EQ(
              src_tensor.dims()[i], dst_tensor->dims()[i],
              platform::errors::InvalidArgument(
                  "`src` and `dst` should have the same tensor shape, "
                  "except for the first dimension."));
        }

        auto stream = paddle::platform::stream::get_current_stream(deviceId)
                          ->raw_stream();

        int64_t size = src_tensor.numel() / src_tensor.dims()[0];
        auto *src_data = src_tensor.data<float>();
        auto *dst_data = dst_tensor->mutable_data<float>(dst.Place());
        const int64_t *offset_data = offset_tensor.data<int64_t>();
        const int64_t *count_data = count_tensor.data<int64_t>();
        int64_t src_offset = 0, dst_offset, c;
        for (int64_t i = 0; i < offset_tensor.numel(); i++) {
          dst_offset = offset_data[i], c = count_data[i];
          PADDLE_ENFORCE_LE(src_offset + c, src_tensor.dims()[0],
                            platform::errors::InvalidArgument(
                                "Invalid offset or count index"));
          PADDLE_ENFORCE_LE(dst_offset + c, dst_tensor->dims()[0],
                            platform::errors::InvalidArgument(
                                "Invalid offset or count index"));
          cudaMemcpyAsync(
              dst_data + (dst_offset * size), src_data + (src_offset * size),
              c * size * sizeof(float), cudaMemcpyDeviceToHost, stream);
          src_offset += c;
        }
      },
      R"DOC(
  This api provides a way to write pieces of source tensor to destination tensor 
  inplacely and asynchronously. In which, we use `offset` and `count` to determine 
  where to copy. `offset` means the begin points of the copy pieces of `src`, and 
  `count` means the lengths of the copy pieces of `src`. To be noted, the copy process 
  will run asynchronously from cuda to pin memory. We can simply remember this as 
  "gpu async_write to pin_memory".
  
  Arguments:
  
    src (Tensor): The source tensor, and the data type should be `float32` currently. 
                  Besides, `src` should be placed on CUDAPlace.

    dst (Tensor): The destination tensor, and the data type should be `float32` currently. 
                  Besides, `dst` should be placed on CUDAPinnedPlace. The shape of `dst` 
                  should be the same with `src` except for the first dimension. 

    offset (Tensor): The offset tensor, and the data type should be `int64` currently. 
                     Besides, `offset` should be placed on CPUPlace. The shape of `offset` 
                     should be one-dimensional. 
    
    count (Tensor): The count tensor, and the data type should be `int64` currently. 
                    Besides, `count` should be placed on CPUPlace. The shape of `count` 
                    should be one-dimensinal. 

  Examples:
      .. code-block:: python

          import numpy as np
          import paddle
          from paddle.fluid import core  
          from paddle.device import cuda
          
          if core.is_compiled_with_cuda():
              src = paddle.rand(shape=[100, 50, 50])
              dst = paddle.emtpy(shape=[200, 50, 50]).pin_memory()
              offset = paddle.to_tensor(
                  np.array([0, 60], dtype="int64"), place=paddle.CPUPlace())
              count = paddle.to_tensor(
                  np.array([40, 60], dtype="int64"), place=paddle.CPUPlace())

              stream = cuda.Stream()
              with cuda.stream_guard(stream):
                  core.async_write(src, dst, offset, count)

              offset_a = paddle.gather(dst, paddle.to_tensor(np.arange(0, 40)))
              offset_b = paddle.gather(dst, paddle.to_tensor(np.arange(60, 120)))
              offset_array = paddle.concat([offset_a, offset_b], axis=0)
              print(np.allclose(src.numpy(), offset_array.numpy())) # True
)DOC");

  m.def(
      "async_read",
      [](const imperative::VarBase &src, imperative::VarBase &dst,
         const imperative::VarBase &index, imperative::VarBase &buffer,
         const imperative::VarBase &offset, const imperative::VarBase &count) {
        PADDLE_ENFORCE_EQ(platform::is_cuda_pinned_place(src.Place()), true,
                          platform::errors::InvalidArgument(
                              "Required `src` device should be "
                              "CUDAPinnedPlace, but received %d.",
                              src.Place()));
        PADDLE_ENFORCE_EQ(
            platform::is_gpu_place(dst.Place()), true,
            platform::errors::InvalidArgument(
                "Required `dst` device should be CUDAPlace, but received %d.",
                dst.Place()));
        PADDLE_ENFORCE_EQ(
            platform::is_cpu_place(index.Place()), true,
            platform::errors::InvalidArgument(
                "Required `index` device should be CPUPlace, but received %d.",
                index.Place()));
        PADDLE_ENFORCE_EQ(
            platform::is_cuda_pinned_place(buffer.Place()), true,
            platform::errors::InvalidArgument(
                "Required `buffer` device should be CUDAPinnedPlace, "
                "but received %d.",
                buffer.Place()));
        PADDLE_ENFORCE_EQ(
            platform::is_cpu_place(offset.Place()), true,
            platform::errors::InvalidArgument(
                "Required `offset` device should be CPUPlace, but received %d.",
                offset.Place()));
        PADDLE_ENFORCE_EQ(
            platform::is_cpu_place(count.Place()), true,
            platform::errors::InvalidArgument(
                "Required `count` device should be CPUPlace, but received %d.",
                count.Place()));

        auto &src_tensor = src.Var().Get<framework::LoDTensor>();
        auto *dst_tensor = dst.MutableVar()->GetMutable<framework::LoDTensor>();
        auto &index_tensor = index.Var().Get<framework::LoDTensor>();
        auto *buffer_tensor =
            buffer.MutableVar()->GetMutable<framework::LoDTensor>();
        auto &offset_tensor = offset.Var().Get<framework::LoDTensor>();
        auto &count_tensor = count.Var().Get<framework::LoDTensor>();
        auto *dst_data = dst_tensor->mutable_data<float>(dst.Place());
        const auto &deviceId = paddle::platform::GetCurrentDeviceId();

        PADDLE_ENFORCE_EQ(src_tensor.dims().size(), dst_tensor->dims().size(),
                          platform::errors::InvalidArgument(
                              "`src` and `dst` should have same tensor shape, "
                              "except for the first dimension."));
        PADDLE_ENFORCE_EQ(
            src_tensor.dims().size(), buffer_tensor->dims().size(),
            platform::errors::InvalidArgument(
                "`src` and `buffer` should have same tensor shape, "
                "except for the first dimension."));
        for (int i = 1; i < src_tensor.dims().size(); i++) {
          PADDLE_ENFORCE_EQ(
              src_tensor.dims()[i], dst_tensor->dims()[i],
              platform::errors::InvalidArgument(
                  "`src` and `dst` should have the same tensor shape, "
                  "except for the first dimension."));
          PADDLE_ENFORCE_EQ(
              src_tensor.dims()[i], buffer_tensor->dims()[i],
              platform::errors::InvalidArgument(
                  "`src` and `buffer` should have the same tensor shape, "
                  "except for the first dimension."));
        }
        PADDLE_ENFORCE_EQ(index_tensor.dims().size(), 1,
                          platform::errors::InvalidArgument(
                              "`index` tensor should be one-dimensional."));

        auto stream = paddle::platform::stream::get_current_stream(deviceId)
                          ->raw_stream();

        int64_t numel = 0;  // total copy length
        int64_t copy_flag = offset_tensor.dims()[0];
        int64_t size = src_tensor.numel() / src_tensor.dims()[0];

        if (copy_flag != 0) {
          PADDLE_ENFORCE_EQ(offset_tensor.dims().size(), 1,
                            platform::errors::InvalidArgument(
                                "`offset` tensor should be one-dimensional."));
          PADDLE_ENFORCE_EQ(count_tensor.dims().size(), 1,
                            platform::errors::InvalidArgument(
                                "`count` tensor should be one-dimensional."));
          PADDLE_ENFORCE_EQ(offset_tensor.numel(), count_tensor.numel(),
                            platform::errors::InvalidArgument(
                                "`offset` and `count` tensor size dismatch."));
          auto *offset_data = offset_tensor.data<int64_t>();
          auto *count_data = count_tensor.data<int64_t>();
          for (int64_t i = 0; i < count_tensor.numel(); i++) {
            numel += count_data[i];
          }
          PADDLE_ENFORCE_LE(numel + index_tensor.numel(),
                            buffer_tensor->dims()[0],
                            platform::errors::InvalidArgument(
                                "Buffer tensor size is too small."));
          PADDLE_ENFORCE_LE(numel + index_tensor.numel(), dst_tensor->dims()[0],
                            platform::errors::InvalidArgument(
                                "Target tensor size is too small."));

          int64_t src_offset, dst_offset = 0, c;
          auto *src_data = src_tensor.data<float>();
          for (int64_t i = 0; i < offset_tensor.numel(); i++) {
            src_offset = offset_data[i], c = count_data[i];
            PADDLE_ENFORCE_LE(src_offset + c, src_tensor.dims()[0],
                              platform::errors::InvalidArgument(
                                  "Invalid offset or count index."));
            PADDLE_ENFORCE_LE(dst_offset + c, dst_tensor->dims()[0],
                              platform::errors::InvalidArgument(
                                  "Invalid offset or count index."));
            cudaMemcpyAsync(
                dst_data + (dst_offset * size), src_data + (src_offset * size),
                c * size * sizeof(float), cudaMemcpyHostToDevice, stream);
            dst_offset += c;
          }
        } else {
          PADDLE_ENFORCE_LE(index_tensor.numel(), buffer_tensor->dims()[0],
                            platform::errors::InvalidArgument(
                                "Buffer tensor size is too small."));
        }

        // Select the index data to the buffer
        auto index_select = [](const framework::Tensor &src_tensor,
                               const framework::Tensor &index_tensor,
                               framework::Tensor *buffer_tensor) {
          auto *src_data = src_tensor.data<float>();
          auto *index_data = index_tensor.data<int64_t>();
          auto *buffer_data =
              buffer_tensor->mutable_data<float>(buffer_tensor->place());
          const int &slice_size = src_tensor.numel() / src_tensor.dims()[0];
          const int &copy_bytes = slice_size * sizeof(float);
          int64_t c = 0;
          for (int64_t i = 0; i < index_tensor.numel(); i++) {
            std::memcpy(buffer_data + c * slice_size,
                        src_data + index_data[i] * slice_size, copy_bytes);
            c += 1;
          }
        };
        index_select(src_tensor, index_tensor, buffer_tensor);

        // Copy the data to device memory
        cudaMemcpyAsync(dst_data + (numel * size), buffer_tensor->data<float>(),
                        index_tensor.numel() * size * sizeof(float),
                        cudaMemcpyHostToDevice, stream);
      },
      R"DOC(
  This api provides a way to read from pieces of source tensor to destination tensor 
  asynchronously. In which, we use `index`, `offset` and `count` to determine where 
  to read. `index` means the index position of src tensor we want to read. `offset` 
  and count means the begin points and length of pieces of src tensor we want to read. 
  To be noted, the copy process will run asynchronously from pin memory to cuda place. 
  We can simply remember this as "cuda async_read from pin_memory".

  Arguments:
  
    src (Tensor): The source tensor, and the data type should be `float32` currently. 
                  Besides, `src` should be placed on CUDAPinnedPlace.
  
    dst (Tensor): The destination tensor, and the data type should be `float32` currently. 
                  Besides, `dst` should be placed on CUDAPlace. The shape of `dst` should 
                  be the same with `src` except for the first dimension.

    index (Tensor): The index tensor, and the data type should be `int64` currently. 
                    Besides, `index` should be on CPUplace. The shape of `index` should 
                    be one-dimensional.

    buffer (Tensor): The buffer tensor, used to buffer index copy tensor temporarily. 
                     The data type should be `float32` currently, and should be placed 
                     on CUDAPinnedPlace. The shape of `buffer` should be the same with `src` except for the first dimension.

    offset (Tensor): The offset tensor, and the data type should be `int64` currently. 
                     Besides, `offset` should be placed on CPUPlace. The shape of `offset` 
                     should be one-dimensional.

    count (Tensor): The count tensor, and the data type should be `int64` currently. 
                    Besides, `count` should be placed on CPUPlace. The shape of `count` 
                    should be one-dimensinal.
    
  Examples:
      .. code-block:: python

          import numpy as np
          import paddle
          from paddle.fluid import core
          from paddle.device import cuda

          if core.is_compiled_with_cuda():
              src = paddle.rand(shape=[100, 50, 50], dtype="float32").pin_memory()
              dst = paddle.empty(shape=[100, 50, 50], dtype="float32")
              offset = paddle.to_tensor(
                  np.array([0, 60], dtype="int64"), place=paddle.CPUPlace())
              count = paddle.to_tensor(
                  np.array([40, 60], dtype="int64"), place=paddle.CPUPlace())
              buffer = paddle.empty(shape=[50, 50, 50], dtype="float32").pin_memory()
              index = paddle.to_tensor(
                  np.array([1, 3, 5, 7, 9], dtype="int64")).cpu()
          
              stream = cuda.Stream()
              with cuda.stream_guard(stream):
                  core.async_read(src, dst, index, buffer, offset, count)
 
)DOC");
#endif
2704 2705 2706 2707
}

}  // namespace pybind
}  // namespace paddle