imperative.cc 41.1 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 <memory>
24
#include <set>
J
Jiabin Yang 已提交
25
#include <string>
26
#include <unordered_map>
27
#include <unordered_set>
28
#include <utility>
J
Jiabin Yang 已提交
29
#include <vector>
30

31
#include "paddle/fluid/imperative/all_reduce.h"
32
#include "paddle/fluid/imperative/amp_auto_cast.h"
J
Jiabin Yang 已提交
33
#include "paddle/fluid/imperative/backward_strategy.h"
34
#include "paddle/fluid/imperative/basic_engine.h"
35
#include "paddle/fluid/imperative/data_loader.h"
36
#include "paddle/fluid/imperative/layer.h"
J
Jiabin Yang 已提交
37
#include "paddle/fluid/imperative/nccl_context.h"
38
#include "paddle/fluid/imperative/partial_grad_engine.h"
39
#include "paddle/fluid/imperative/profiler.h"
40
#include "paddle/fluid/imperative/tracer.h"
M
minqiyang 已提交
41
#include "paddle/fluid/imperative/type_defs.h"
42
#include "paddle/fluid/memory/allocation/mmap_allocator.h"
43
#include "paddle/fluid/pybind/op_function.h"
44
#include "paddle/fluid/pybind/pybind_boost_headers.h"
L
Leo Chen 已提交
45
#include "paddle/fluid/pybind/tensor_py.h"
46

47 48 49
namespace paddle {
namespace pybind {

50 51
namespace py = ::pybind11;

52 53 54 55
class Layer : public imperative::Layer {
 public:
  using imperative::Layer::Layer;  // Inherit constructors

56 57 58 59
  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
60
                      Forward, inputs);  // NOLINT
61 62 63
  }
};

L
Leo Chen 已提交
64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86
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>();
  } else if (py::isinstance<platform::CUDAPinnedPlace>(place_obj)) {
    return place_obj.cast<platform::CUDAPinnedPlace>();
  } else {
    PADDLE_THROW(platform::errors::InvalidArgument(
        "Place should be one of CPUPlace/CUDAPlace/CUDAPinnedPlace"));
  }
}

static void InitTensorForVarBase(imperative::VarBase *self,
                                 const py::array &array,
                                 const platform::Place place,
                                 bool persistable = false,
                                 bool zero_copy = false,
                                 std::string name = "") {
  if (name == "") {
    name = imperative::GetCurrentTracer()->GenerateUniqueName("generated_var");
  }
  new (self) imperative::VarBase(name);
87
  auto *tensor = self->MutableVar()->GetMutable<framework::LoDTensor>();
L
Leo Chen 已提交
88 89
  if (platform::is_cpu_place(place)) {
    SetTensorFromPyArray<platform::CPUPlace>(
90
        tensor, array, BOOST_GET_CONST(platform::CPUPlace, place), zero_copy);
L
Leo Chen 已提交
91 92
  } else if (platform::is_gpu_place(place)) {
    SetTensorFromPyArray<platform::CUDAPlace>(
93
        tensor, array, BOOST_GET_CONST(platform::CUDAPlace, place), zero_copy);
L
Leo Chen 已提交
94 95
  } else if (platform::is_cuda_pinned_place(place)) {
    SetTensorFromPyArray<platform::CUDAPinnedPlace>(
96 97
        tensor, array, BOOST_GET_CONST(platform::CUDAPinnedPlace, place),
        zero_copy);
98
  } else {
L
Leo Chen 已提交
99 100
    PADDLE_THROW(platform::errors::InvalidArgument(
        "Place should be one of CPUPlace/CUDAPlace/CUDAPinnedPlace"));
J
Jiabin Yang 已提交
101
  }
L
Leo Chen 已提交
102
  self->SetPersistable(persistable);
103 104 105 106 107 108
  self->SetType(framework::proto::VarType::LOD_TENSOR);
  self->SetDataType(tensor->type());
}

static void InitVarBaseFromNumpyWithKwargs(imperative::VarBase *self,
                                           const py::kwargs &kwargs) {
109
  VLOG(4) << "Init VarBase";
110 111
  PADDLE_ENFORCE_EQ(
      kwargs.contains("value"), true,
112 113
      platform::errors::NotFound(
          "The kwargs used to create Varbase misses argument: value"));
L
Leo Chen 已提交
114 115 116 117 118 119 120 121 122 123 124 125 126

  auto persistable = kwargs.contains("persistable")
                         ? kwargs["persistable"].cast<bool>()
                         : false;
  auto array = kwargs.contains("value") ? kwargs["value"].cast<py::array>()
                                        : py::array();
  auto zero_copy =
      kwargs.contains("zero_copy") ? kwargs["zero_copy"].cast<bool>() : false;
  auto name = kwargs.contains("name") ? kwargs["name"].cast<std::string>() : "";
  auto default_place = imperative::GetCurrentTracer()->ExpectedPlace();
  auto place = kwargs.contains("place") ? PyObjectToPlace(kwargs["place"])
                                        : default_place;
  InitTensorForVarBase(self, array, place, persistable, zero_copy, name);
127
}
128

129 130 131
template <typename P>
static void InitVarBaseFromNumpyWithArg(imperative::VarBase *self,
                                        const py::array &array, const P &place,
L
Leo Chen 已提交
132 133 134
                                        bool persistable = false,
                                        bool zero_copy = false,
                                        std::string name = "") {
135
  VLOG(4) << "Init VarBase";
L
Leo Chen 已提交
136 137 138 139 140
  // 0: self, 1: value, 2: place, 3: persistable, 4: zero_copy, 5: name
  if (name == "") {
    name = imperative::GetCurrentTracer()->GenerateUniqueName("generated_var");
  }
  new (self) imperative::VarBase(name);
141 142 143 144 145 146 147 148
  self->SetPersistable(persistable);
  auto *tensor = self->MutableVar()->GetMutable<framework::LoDTensor>();
  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
149
                                               const py::array &array) {
150
  VLOG(4) << "Init VarBase";
L
Leo Chen 已提交
151 152
  auto place = imperative::GetCurrentTracer()->ExpectedPlace();
  InitTensorForVarBase(self, array, place);
153
}
154

155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174
static void InitVarBaseFromTensorWithArgDefault(
    imperative::VarBase *self, const framework::LoDTensor &tensor) {
  VLOG(4) << "Init VarBase";
  auto place = imperative::GetCurrentTracer()->ExpectedPlace();
  new (self) imperative::VarBase(
      imperative::GetCurrentTracer()->GenerateUniqueName("generated_var"));
  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";
  }
}

175 176 177 178 179
static std::string GetTypeName(const imperative::VarBase &var) {
  if (var.Type() == framework::proto::VarType::RAW) {
    return "RAW";
  } else if (!var.Var().IsInitialized()) {
    return "nullptr";
180
  } else {
181
    return framework::ToTypeName(var.Var().Type());
182 183
  }
}
L
Leo Chen 已提交
184

185
using PyNameVarBaseMap = std::unordered_map<std::string, py::handle>;
186 187 188 189 190 191

template <typename T>
static T PyObjectCast(PyObject *obj) {
  try {
    return py::cast<T>(py::handle(obj));
  } catch (py::cast_error &) {
192 193
    PADDLE_THROW(platform::errors::InvalidArgument(
        "Python object is not type of %s", typeid(T).name()));
194 195 196 197 198 199 200 201 202 203 204 205 206 207 208
  }
}

// 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;

209
  if (PyList_Check(py_obj)) {  // List of VarBase
210 211 212
    size_t len = PyList_GET_SIZE(py_obj);
    result.reserve(len);
    for (size_t i = 0; i < len; ++i) {
213 214 215
      PyObject *py_ivar = PyList_GET_ITEM(py_obj, i);
      PADDLE_ENFORCE_NOT_NULL(
          py_ivar, platform::errors::InvalidArgument("Python Object is NULL"));
216 217 218
      result.emplace_back(
          PyObjectCast<std::shared_ptr<imperative::VarBase>>(py_ivar));
    }
219
  } else if (PyTuple_Check(py_obj)) {  // Tuple of VarBase
220 221 222
    size_t len = PyTuple_GET_SIZE(py_obj);
    result.reserve(len);
    for (size_t i = 0; i < len; ++i) {
223 224 225
      PyObject *py_ivar = PyTuple_GET_ITEM(py_obj, i);
      PADDLE_ENFORCE_NOT_NULL(
          py_ivar, platform::errors::InvalidArgument("Python Object is NULL"));
226 227 228
      result.emplace_back(
          PyObjectCast<std::shared_ptr<imperative::VarBase>>(py_ivar));
    }
229 230 231
  } else {  // VarBase
    result.emplace_back(
        PyObjectCast<std::shared_ptr<imperative::VarBase>>(py_obj));
232 233 234 235 236
  }

  return result;
}

J
Jiabin Yang 已提交
237 238 239
static imperative::NameVarBaseMap ConvertToNameVarBaseMap(
    const PyNameVarBaseMap &map) {
  imperative::NameVarBaseMap result;
240 241 242 243 244 245
  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
246

247 248 249
  PADDLE_ENFORCE_EQ(
      PyErr_Occurred(), nullptr,
      platform::errors::InvalidArgument(py::str(py::handle(PyErr_Occurred()))));
250 251 252
  return result;
}

253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317
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
}

// 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 {
    if (PyCheckInteger(r->step)) {
      *step = PyLong_AsLong(r->step);
    } else {
      PADDLE_THROW(platform::errors::InvalidArgument(
          "Currently, VarBase.__getitem__() only allows None or integers in "
          "slice item, but received %s.",
          std::string(Py_TYPE(r->step)->tp_name)));
    }
  }
  if (r->start == Py_None) {
    *start = *step < 0 ? length - 1 : 0;
  } else {
    if (PyCheckInteger(r->start)) {
      *start = PyLong_AsLong(r->start);
    } else {
      PADDLE_THROW(platform::errors::InvalidArgument(
          "Currently, VarBase.__getitem__() only allows None or integers in "
          "slice item, but received %s.",
          std::string(Py_TYPE(r->start)->tp_name)));
    }
    if (*start < 0) *start += length;
  }
  if (r->stop == Py_None) {
    *stop = *step < 0 ? -1 : length;
  } else {
    if (PyCheckInteger(r->stop)) {
      *stop = PyLong_AsLong(r->stop);
    } else {
      PADDLE_THROW(platform::errors::InvalidArgument(
          "Currently, VarBase.__getitem__() only allows None or integers in "
          "slice item, but received %s.",
          std::string(Py_TYPE(r->stop)->tp_name)));
    }
    if (*stop < 0) *stop += length;
  }
  if (*stop > length) return -1;
  if (*start >= length) return -1;
  if (*step == 0) return -1;
  return 0;
}

S
songyouwei 已提交
318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341
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> *infer_flags) {
  // We allow indexing by Integers, Slices, and tuples of those
  // types.
  // Ellipsis and None are not supported yet.
  // wrap to tuple
  PyObject *index = !PyTuple_Check(_index) ? PyTuple_Pack(1, _index) : _index;
  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);
  PADDLE_ENFORCE_EQ(
      size <= rank, true,
      platform::errors::InvalidArgument(
          "too many indices (%d) for tensor of dimension %d", size, rank));
  for (int dim = 0; dim < size; ++dim) {
    PyObject *slice_item = PyTuple_GetItem(index, dim);
342 343 344 345 346 347 348
    PADDLE_ENFORCE_EQ(PyCheckInteger(slice_item) || PySlice_Check(slice_item),
                      true,
                      platform::errors::InvalidArgument(
                          "Currently, VarBase.__getitem__() only allows "
                          "indexing by Integers, Slices, and tuples of "
                          "these types, but received %s in %dth slice item",
                          std::string(Py_TYPE(slice_item)->tp_name), dim + 1));
S
songyouwei 已提交
349 350
    infer_flags->push_back(1);
    int dim_len = shape[dim];
351 352
    if (PyCheckInteger(slice_item)) {
      // integer, PyLong_AsLong supports both int and long
S
songyouwei 已提交
353
      int start = static_cast<int>(PyLong_AsLong(slice_item));
H
hong 已提交
354
      auto s_t = start;
S
songyouwei 已提交
355
      start = start < 0 ? start + dim_len : start;
H
hong 已提交
356 357 358 359 360 361 362 363 364 365
      if (start >= dim_len) {
        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
366 367 368 369 370 371
      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);
    } else {
372
      // slice item
S
songyouwei 已提交
373
      Py_ssize_t start, end, step;
374 375 376
      PySliceObject *p = reinterpret_cast<PySliceObject *>(slice_item);
      _PySlice_GetIndices(p, dim_len, &start, &end, &step);

S
songyouwei 已提交
377
      // :: or : or 0:dim_len:1
378 379 380
      if (start == 0 && end == dim_len && step == 1) {
        continue;
      }
S
songyouwei 已提交
381 382 383 384 385 386 387 388 389
      slice_axes->push_back(dim);
      slice_starts->push_back(start);
      slice_ends->push_back(end);
      slice_strides->push_back(step);
    }
  }
  if (!PyTuple_Check(_index)) Py_DecRef(index);
}

390
// Bind Methods
J
Jiabin Yang 已提交
391
void BindImperative(py::module *m_ptr) {
392 393
  auto &m = *m_ptr;

394 395
  BindOpFunctions(&m);

396 397
#ifndef _WIN32
  // Dygraph DataLoader signal handler
398 399 400 401 402 403 404 405 406 407 408 409 410
  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);
411
  });
412 413
  m.def("_erase_process_pids",
        [](int64_t key) { imperative::EraseLoadProcessPIDs(key); });
414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485
  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);

  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

486
  py::class_<imperative::detail::BackwardStrategy> backward_strategy(
487 488
      m, "BackwardStrategy", R"DOC(

J
Jiabin Yang 已提交
489
    BackwardStrategy is a descriptor of how to run the backward process.
490

J
Jiabin Yang 已提交
491
    **Note**:
T
tianshuo78520a 已提交
492
        **This API is only available in** `Dygraph <../../user_guides/howto/dygraph/DyGraph.html>`_ **Mode**
493

J
Jiabin Yang 已提交
494 495
    Attribute:
        **sort_sum_gradient**:
496

J
Jiabin Yang 已提交
497
        If framework will sum the gradient by the reverse order of trace. eg. x_var ( :ref:`api_guide_Variable` ) will be the input of multiple OP such as :ref:`api_fluid_layers_scale` , this attr will decide if framework will sum gradient of `x_var` by the reverse order.
L
lujun 已提交
498

J
Jiabin Yang 已提交
499
        By Default: False
L
lujun 已提交
500

J
Jiabin Yang 已提交
501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518
        Examples:
            .. code-block:: python

                import numpy as np
                import paddle.fluid as fluid

                x = np.ones([2, 2], np.float32)
                with fluid.dygraph.guard():
                    x_var = fluid.dygraph.to_variable(x)
                    sums_inputs = []
                    # x_var will be multi-scales' input here
                    for _ in range(10):
                        sums_inputs.append(fluid.layers.scale(x_var))
                    ret2 = fluid.layers.sums(sums_inputs)
                    loss2 = fluid.layers.reduce_sum(ret2)
                    backward_strategy = fluid.dygraph.BackwardStrategy()
                    backward_strategy.sort_sum_gradient = True
                    loss2.backward(backward_strategy)
519
      )DOC");
520 521 522 523 524 525 526 527 528 529 530 531 532 533 534
  backward_strategy.def(py::init())
      .def_property("sort_sum_gradient",
                    [](const imperative::detail::BackwardStrategy &self) {
                      return self.sorted_sum_gradient_;
                    },
                    [](imperative::detail::BackwardStrategy &self,
                       bool sorted_sum_gradient) {
                      self.sorted_sum_gradient_ = sorted_sum_gradient;
                    });

  m.def("start_imperative_gperf_profiler",
        []() { imperative::StartProfile(); });

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

Z
Zeng Jinle 已提交
535 536 537
  m.def("_is_dygraph_debug_enabled",
        []() { return imperative::IsDebugEnabled(); });
  m.def("_dygraph_debug_level", []() { return imperative::GetDebugLevel(); });
538 539 540 541
  m.def("_switch_tracer",
        [](const std::shared_ptr<imperative::Tracer> &tracer) {
          imperative::SetCurrentTracer(tracer);
        });
Z
Zeng Jinle 已提交
542

543
  py::class_<imperative::VarBase, std::shared_ptr<imperative::VarBase>>(
544
      m, "VarBase", R"DOC()DOC")
Z
Zeng Jinle 已提交
545
      .def_static("_alive_vars", &imperative::VarBase::AliveVarNames)
J
Jiabin Yang 已提交
546
      .def("__init__",
547 548 549
           [](imperative::VarBase &self, framework::proto::VarType::Type dtype,
              const std::vector<int> &dims, const py::handle &name,
              framework::proto::VarType::Type type, bool persistable) {
550
             VLOG(4) << "Init VarBase";
551 552 553 554 555 556 557 558
             std::string act_name = "";
             if (!name.ptr() || name.ptr() == Py_None) {
               act_name = imperative::GetCurrentTracer()->GenerateUniqueName(
                   "generated_var");
             } else {
               act_name = name.cast<std::string>();
             }
             new (&self) imperative::VarBase(act_name);
J
Jiabin Yang 已提交
559 560 561 562 563 564 565 566 567
             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));
             }
           })
568 569
      .def("__init__", &InitVarBaseFromNumpyWithArg<platform::CPUPlace>,
           py::arg("value"), py::arg("place"), py::arg("persistable") = false,
L
Leo Chen 已提交
570
           py::arg("zero_copy") = false, py::arg("name") = "")
571 572
      .def("__init__", &InitVarBaseFromNumpyWithArg<platform::CUDAPlace>,
           py::arg("value"), py::arg("place"), py::arg("persistable") = false,
L
Leo Chen 已提交
573
           py::arg("zero_copy") = false, py::arg("name") = "")
574 575
      .def("__init__", &InitVarBaseFromNumpyWithArg<platform::CUDAPinnedPlace>,
           py::arg("value"), py::arg("place"), py::arg("persistable") = false,
L
Leo Chen 已提交
576 577
           py::arg("zero_copy") = false, py::arg("name") = "")
      .def("__init__", &InitVarBaseFromNumpyWithArgDefault, py::arg("value"))
578
      .def("__init__", &InitVarBaseFromTensorWithArgDefault, py::arg("tensor"))
579
      .def("__init__", &InitVarBaseFromNumpyWithKwargs)
580
      .def("__getitem__",
S
songyouwei 已提交
581
           [](std::shared_ptr<imperative::VarBase> &self, py::handle _index) {
582
             std::vector<int> slice_axes, slice_starts, slice_ends,
S
songyouwei 已提交
583 584 585 586 587 588
                 slice_strides, decrease_axis, infer_flags;
             auto tensor =
                 self->MutableVar()->GetMutable<framework::LoDTensor>();
             ParseIndexingSlice(tensor, _index.ptr(), &slice_axes,
                                &slice_starts, &slice_ends, &slice_strides,
                                &decrease_axis, &infer_flags);
589 590 591 592
             // release gil and do tracing
             py::gil_scoped_release release;
             const auto &tracer = imperative::GetCurrentTracer();
             if (slice_axes.empty()) {
S
songyouwei 已提交
593
               return self;
594
             } else {
S
songyouwei 已提交
595
               imperative::NameVarBaseMap ins = {{"Input", {self}}};
596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617
               framework::AttributeMap attrs = {
                   {"axes", slice_axes},
                   {"starts", slice_starts},
                   {"ends", slice_ends},
                   {"infer_flags", infer_flags},
                   {"decrease_axis", decrease_axis}};
               auto out = std::shared_ptr<imperative::VarBase>(
                   new imperative::VarBase(tracer->GenerateUniqueName()));
               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));
               return out;
             }
           })
618 619 620 621 622 623 624
      .def("numpy",
           [](imperative::VarBase &self) -> py::array {
             const auto &tensor =
                 self.MutableVar()->Get<framework::LoDTensor>();
             PADDLE_ENFORCE_EQ(
                 tensor.IsInitialized(), true,
                 platform::errors::InvalidArgument(
625
                     "Tensor of %s is Empty, please check if it has no data.",
626 627 628 629 630
                     self.Name()));
             return TensorToPyArray(tensor, true);
           },
           R"DOC(
        **Notes**:
T
tianshuo78520a 已提交
631
            **This API is ONLY available in Dygraph mode**
632 633 634 635 636 637 638 639 640 641 642 643 644 645

        Returns a numpy array shows the value of current :ref:`api_guide_Variable_en`

        Returns:
            ndarray: The numpy value of current Variable.

        Returns type:
            ndarray: dtype is same as current Variable

        Examples:
            .. code-block:: python

                import paddle.fluid as fluid
                from paddle.fluid.dygraph.base import to_variable
646
                from paddle.fluid.dygraph import Linear
647 648 649 650
                import numpy as np

                data = np.random.uniform(-1, 1, [30, 10, 32]).astype('float32')
                with fluid.dygraph.guard():
651
                    linear = Linear(32, 64)
652
                    data = to_variable(data)
653
                    x = linear(data)
654 655 656 657 658 659 660 661 662 663 664 665 666
                    print(x.numpy())

       )DOC")
      .def("detach",
           [](const 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()));
             return self.NewVarBase(tensor.place(), false);
           },
           py::return_value_policy::copy, R"DOC(
        **Notes**:
T
tianshuo78520a 已提交
667
            **This API is ONLY available in Dygraph mode**
668 669 670 671 672 673 674 675 676 677 678 679

        Returns a new Variable, detached from the current graph.

        Returns:
             ( :ref:`api_guide_Variable_en` | dtype is same as current Variable): The detached Variable.


        Examples:
            .. code-block:: python

                import paddle.fluid as fluid
                from paddle.fluid.dygraph.base import to_variable
680
                from paddle.fluid.dygraph import Linear
681 682 683 684
                import numpy as np

                data = np.random.uniform(-1, 1, [30, 10, 32]).astype('float32')
                with fluid.dygraph.guard():
685
                    linear = Linear(32, 64)
686
                    data = to_variable(data)
687
                    x = linear(data)
688 689 690 691 692 693
                    y = x.detach()

       )DOC")
      .def("clear_gradient", &imperative::VarBase::ClearGradient, R"DOC(

        **Notes**:
T
tianshuo78520a 已提交
694
        **1. This API is ONLY available in Dygraph mode**
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

        **2. Use it only Variable has gradient, normally we use this for Parameters since other temporal Variable will be deleted by Python's GC**

        Clear  (set to ``0`` ) the Gradient of Current Variable

        Returns:  None

        Examples:
             .. code-block:: python

                import paddle.fluid as fluid
                import numpy as np

                x = np.ones([2, 2], np.float32)
                with fluid.dygraph.guard():
                    inputs2 = []
                    for _ in range(10):
                         tmp = fluid.dygraph.base.to_variable(x)
                         tmp.stop_gradient=False
                         inputs2.append(tmp)
                    ret2 = fluid.layers.sums(inputs2)
                    loss2 = fluid.layers.reduce_sum(ret2)
                    backward_strategy = fluid.dygraph.BackwardStrategy()
                    backward_strategy.sort_sum_gradient = True
                    loss2.backward(backward_strategy)
                    print(loss2.gradient())
                    loss2.clear_gradient()
                    print("After clear {}".format(loss2.gradient()))
      )DOC")
L
Leo Chen 已提交
724 725 726
      .def("_run_backward",
           [](imperative::VarBase &self,
              const imperative::detail::BackwardStrategy &bckst,
727
              const imperative::Tracer &tracer, bool retain_graph) {
728 729
             // TODO(jiabin): when we impl more backward execution we can
             // select them
730
             auto *engine = tracer.GetEngine();
731
             engine->Init(&self, bckst, retain_graph);
732
             VLOG(3) << "Start backward";
L
Leo Chen 已提交
733 734 735 736 737 738 739 740 741 742
             engine->Execute();
             VLOG(3) << "Finish backward";
           },
           py::call_guard<py::gil_scoped_release>())
      .def("_grad_name", &imperative::VarBase::GradVarName)
      .def("_grad_value",
           [](imperative::VarBase &self) {
             return self.MutableGradVar()->Get<framework::LoDTensor>();
           },
           py::return_value_policy::reference)
743 744 745 746
      .def("_set_grad_type",
           [](imperative::VarBase &self, framework::proto::VarType::Type type) {
             self.MutableGradVarBase()->SetType(type);
           })
747
      .def("_grad_ivar",
J
Jiabin Yang 已提交
748 749
           [](const imperative::VarBase &self) {
             auto &grad_var = self.GradVarBase();
750 751 752 753 754 755 756 757 758 759 760
             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
761
             }
762
             return std::shared_ptr<imperative::VarBase>(nullptr);
J
Jiabin Yang 已提交
763 764
           },
           py::return_value_policy::copy)
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
      .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) {
#ifdef PADDLE_WITH_NCCL
#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."));
#endif  // PADDLE_WITH_NCCL
             }
           },
           py::call_guard<py::gil_scoped_release>())
795 796
      .def("_copy_to",
           [](const imperative::VarBase &self, const platform::CPUPlace &place,
J
Jiabin Yang 已提交
797 798
              bool blocking) { return self.NewVarBase(place, blocking); },
           py::return_value_policy::copy)
799 800
      .def("_copy_to",
           [](const imperative::VarBase &self, const platform::CUDAPlace &place,
J
Jiabin Yang 已提交
801 802 803
              bool blocking) { return self.NewVarBase(place, blocking); },
           py::return_value_policy::copy)
      .def("value", [](imperative::VarBase &self) { return self.MutableVar(); },
804 805 806
           py::return_value_policy::reference)
      .def_property("name", &imperative::VarBase::Name,
                    &imperative::VarBase::SetName)
L
Leo Chen 已提交
807 808 809 810 811
      .def_property("stop_gradient",
                    &imperative::VarBase::OverridedStopGradient,
                    &imperative::VarBase::SetOverridedStopGradient)
      .def_property("persistable", &imperative::VarBase::Persistable,
                    &imperative::VarBase::SetPersistable)
J
Jiabin Yang 已提交
812 813 814 815
      .def_property_readonly(
          "shape",
          [](imperative::VarBase &self) {
            if (self.Var().IsType<framework::LoDTensor>()) {
816
              return framework::vectorize<int>(
J
Jiabin Yang 已提交
817
                  self.Var().Get<framework::LoDTensor>().dims());
818 819 820
            } else if (self.Var().IsType<framework::SelectedRows>()) {
              return framework::vectorize<int>(
                  self.Var().Get<framework::SelectedRows>().value().dims());
J
Jiabin Yang 已提交
821 822 823 824 825 826 827
            } else {
              VLOG(2) << "It is meaningless to get shape of variable type "
                      << GetTypeName(self);
              return std::vector<int>();
            }
          })
      .def_property_readonly("type", &imperative::VarBase::Type)
L
Leo Chen 已提交
828
      .def_property_readonly("dtype", &imperative::VarBase::DataType);
829 830 831

  py::class_<imperative::Layer, Layer /* <--- trampoline*/> layer(m, "Layer");
  layer.def(py::init<>())
832 833 834 835 836
      .def("forward",
           [](imperative::Layer &self,
              const std::vector<std::shared_ptr<imperative::VarBase>> &inputs) {
             return self.Forward(inputs);
           });
837

838 839 840 841 842
  py::class_<imperative::jit::ProgramDescTracer>(m, "ProgramDescTracer", "")
      .def("create_program_desc",
           &imperative::jit::ProgramDescTracer::CreateProgramDesc)
      .def("reset", &imperative::jit::ProgramDescTracer::Reset);

843
  py::class_<imperative::Tracer, std::shared_ptr<imperative::Tracer>>(
844
      m, "Tracer", R"DOC()DOC")
845
      .def("__init__",
J
Jiabin Yang 已提交
846
           [](imperative::Tracer &self) { new (&self) imperative::Tracer(); })
847 848 849
      .def_property("_enable_program_desc_tracing",
                    &imperative::Tracer::IsProgramDescTracingEnabled,
                    &imperative::Tracer::SetEnableProgramDescTracing)
850 851
      .def_property("_enable_autocast", &imperative::Tracer::IsAutoCastEnabled,
                    &imperative::Tracer::SetEnableAutoCast)
852 853
      .def_property("_train_mode", &imperative::Tracer::HasGrad,
                    &imperative::Tracer::SetHasGrad)
854 855 856 857 858 859 860 861
      .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
862
              self.SetExpectedPlace(*p);
863 864
            } else if (py::isinstance<platform::CPUPlace>(obj)) {
              auto p = obj.cast<platform::CPUPlace *>();
L
Leo Chen 已提交
865
              self.SetExpectedPlace(*p);
866 867
            } else if (py::isinstance<platform::CUDAPinnedPlace>(obj)) {
              auto p = obj.cast<platform::CUDAPinnedPlace *>();
L
Leo Chen 已提交
868
              self.SetExpectedPlace(*p);
869
            } else {
L
Leo Chen 已提交
870
              PADDLE_THROW(platform::errors::InvalidArgument(
871
                  "Incompatible Place Type: supports CUDAPlace, CPUPlace, "
L
Leo Chen 已提交
872 873
                  "and CUDAPinnedPlace, "
                  "but got Unknown Type!"));
874 875
            }
          })
876 877 878
      .def("_get_program_desc_tracer",
           &imperative::Tracer::GetProgramDescTracer,
           py::return_value_policy::reference)
879
      .def("_generate_unique_name", &imperative::Tracer::GenerateUniqueName,
880
           py::arg("key") = "eager_tmp")
881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900
      .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().GetAllowOps()->swap(allow_ops);
            imperative::AmpOperators::Instance().GetBlockOps()->swap(block_ops);
          })
      .def("_get_amp_op_list",
           [](imperative::Tracer &self) {
             return std::make_tuple(
                 *(imperative::AmpOperators::Instance().GetAllowOps()),
                 *(imperative::AmpOperators::Instance().GetBlockOps()));
           })
M
minqiyang 已提交
901
      .def("trace",
J
Jiabin Yang 已提交
902 903 904 905 906 907
           [](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);
908 909
             {
               py::gil_scoped_release release;
J
Jiabin Yang 已提交
910 911
               self.TraceOp(type, std::move(ins_map), std::move(outs_map),
                            std::move(attrs), place, trace_backward);
912
             }
M
minqiyang 已提交
913
           })
J
Jiabin Yang 已提交
914 915 916 917 918 919 920 921 922 923 924 925 926
      .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);
             }
           });
927 928

  // define parallel context
929 930 931
  py::class_<imperative::ParallelStrategy> parallel_strategy(
      m, "ParallelStrategy", "");
  parallel_strategy.def(py::init())
932 933
      .def_property(
          "nranks",
934 935
          [](const imperative::ParallelStrategy &self) { return self.nranks_; },
          [](imperative::ParallelStrategy &self, int nranks) {
936 937 938
            self.nranks_ = nranks;
          })
      .def_property("local_rank",
939
                    [](const imperative::ParallelStrategy &self) {
940 941
                      return self.local_rank_;
                    },
942
                    [](imperative::ParallelStrategy &self, int local_rank) {
943 944 945 946
                      self.local_rank_ = local_rank;
                    })
      .def_property(
          "trainer_endpoints",
947
          [](const imperative::ParallelStrategy &self) {
948 949
            return self.trainer_endpoints_;
          },
950
          [](imperative::ParallelStrategy &self, std::vector<std::string> eps) {
951 952 953
            self.trainer_endpoints_ = eps;
          })
      .def_property("current_endpoint",
954
                    [](const imperative::ParallelStrategy &self) {
955 956
                      return self.current_endpoint_;
                    },
957 958
                    [](imperative::ParallelStrategy &self,
                       const std::string &ep) { self.current_endpoint_ = ep; });
959 960 961 962 963 964 965 966 967 968

  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,
         const platform::Place &place,
         const imperative::detail::BackwardStrategy &strategy,
Z
Zeng Jinle 已提交
969 970 971 972 973
         bool create_graph, bool retain_graph, bool allow_unused,
         bool only_inputs) {
        imperative::PartialGradEngine engine(
            input_targets, output_targets, output_grads, no_grad_vars, place,
            strategy, create_graph, retain_graph, allow_unused, only_inputs);
974 975 976 977 978
        engine.Execute();
        return engine.GetResult();
      },
      py::call_guard<py::gil_scoped_release>());

979
#if defined(PADDLE_WITH_NCCL)
980 981
  py::class_<imperative::NCCLParallelContext> nccl_ctx(m,
                                                       "NCCLParallelContext");
982 983

  nccl_ctx
984 985 986
      .def(py::init<const imperative::ParallelStrategy &,
                    const platform::CUDAPlace &>())
      .def("init", [](imperative::NCCLParallelContext &self) { self.Init(); });
987
#endif
988 989 990 991
}

}  // namespace pybind
}  // namespace paddle