Support numpy scalar parameters (#7935)

* feat(functional): support numpy scalar parameters

* rename inferface

* feat(*): TensorIndex support numpy scalar

* feat(TensorIndex): support advance indexing

* add unittest and int32 support for branch feat-param_support_np_scalar (#7939)

* add unittest

* refactor unittest

* add todo for int16 advanced indexing

* add int32 supporting for advance indexing

* auto format by CI

Co-authored-by: Wang Yi <53533850+marigoold@users.noreply.github.com>
Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
Co-authored-by: oneflow-ci-bot <ci-bot@oneflow.org>

oneflow/api/python/functional/common.cpp

@@ -19,6 +19,7 @@ limitations under the License.
 #include <string>
 
 #include "oneflow/api/python/functional/indexing.h"
+#include "oneflow/extension/python/numpy.h"
 #include "oneflow/core/common/just.h"
 #include "oneflow/core/common/scalar.h"
 #include "oneflow/core/framework/dtype.h"
@@ -195,18 +196,20 @@ Maybe<std::vector<Symbol<SbpParallel>>> PyUnpackSbpParallelSequence(PyObject* ob
 
 // Tensor index
 bool PyTensorIndexCheck(PyObject* obj) {
-  return PySlice_Check(obj) || PyLong_Check(obj) || obj == Py_Ellipsis || obj == Py_None
-         || PyTensorCheck(obj) || PySequence_Check(obj) || PyUnicode_Check(obj);
+  return PySlice_Check(obj) || PyLong_Check(obj) || numpy::PyArrayCheckLongScalar(obj)
+         || obj == Py_Ellipsis || obj == Py_None || PyTensorCheck(obj) || PySequence_Check(obj)
+         || PyUnicode_Check(obj);
 }
 Maybe<TensorIndex> PyUnpackTensorIndex(PyObject* obj) {
   auto tensor_index = std::make_shared<TensorIndex>();
   // Obvious single-entry cases.
-  if (PySlice_Check(obj)         // NOLINT
-      || PyLong_Check(obj)       // NOLINT
-      || obj == Py_Ellipsis      // NOLINT
-      || obj == Py_None          // NOLINT
-      || PyTensorCheck(obj)      // NOLINT
-      || !PySequence_Check(obj)  // NOLINT
+  if (PySlice_Check(obj)                     // NOLINT
+      || PyLong_Check(obj)                   // NOLINT
+      || numpy::PyArrayCheckLongScalar(obj)  // NOLINT
+      || obj == Py_Ellipsis                  // NOLINT
+      || obj == Py_None                      // NOLINT
+      || PyTensorCheck(obj)                  // NOLINT
+      || !PySequence_Check(obj)              // NOLINT
       || PyUnicode_Check(obj)) {
     tensor_index->emplace_back(*JUST(detail::UnpackIndexItem(obj)));
     return tensor_index;

oneflow/api/python/functional/indexing.cpp

@@ -65,6 +65,8 @@ Maybe<DataType> InferScalarType(PyObject* object) {
     return DataType::kInt64;
   } else if (PyArray_Check(object)) {
     return numpy::GetOFDataTypeFromNpArray(reinterpret_cast<PyArrayObject*>(object));
+  } else if (PyArray_CheckScalar(object)) {
+    return numpy::NumpyTypeToOFDataType(PyArray_DescrFromScalar(object)->type_num);
   } else if (PySequence_Check(object)) {
     int64_t length = PySequence_Length(object);
     CHECK_GT_OR_RETURN(length, 0) << "Index should not be empty.";
@@ -86,13 +88,20 @@ Maybe<DataType> InferScalarType(PyObject* object) {
 
 Maybe<void> ParseScalar(PyObject* object, char* data, const DataType& dtype) {
   if (dtype == DataType::kInt64) {
-    CHECK_OR_RETURN(PyLong_Check(object)) << "Expected a long value.";
+    CHECK_OR_RETURN(PyLong_Check(object) || numpy::PyArrayCheckLongScalar(object))
+        << "Expected a long value.";
     *(reinterpret_cast<int64_t*>(data)) = PyLong_AsLongLong(object);
     return Maybe<void>::Ok();
+  } else if (dtype == DataType::kInt32) {
+    CHECK_OR_RETURN(PyLong_Check(object) || numpy::PyArrayCheckLongScalar(object))
+        << "Expected a long value.";
+    *(reinterpret_cast<int32_t*>(data)) = PyLong_AsLongLong(object);
+    return Maybe<void>::Ok();
   } else if (dtype == DataType::kUInt8 || dtype == DataType::kBool) {
-    CHECK_OR_RETURN(PyBool_Check(object) || PyLong_Check(object))
+    CHECK_OR_RETURN(PyBool_Check(object) || PyLong_Check(object)
+                    || numpy::PyArrayCheckLongScalar(object))
         << "Expected a boolean or long value.";
-    if (PyBool_Check(object)) {
+    if (PyBool_Check(object) || numpy::PyArrayCheckBoolScalar(object)) {
       *(reinterpret_cast<bool*>(data)) = (object == Py_True);
     } else {
       int64_t value = PyLong_AsLongLong(object);
@@ -193,6 +202,8 @@ Maybe<IndexItem> UnpackIndexItem(PyObject* object) {
     return std::make_shared<IndexItem>(start, end, step);
   } else if (PyLong_Check(object) && object != Py_False && object != Py_True) {
     return std::make_shared<IndexItem>(static_cast<int64_t>(PyLong_AsLongLong(object)));
+  } else if (numpy::PyArrayCheckLongScalar(object)) {
+    return std::make_shared<IndexItem>(static_cast<int64_t>(PyLong_AsLongLong(object)));
   } else if (object == Py_False || object == Py_True) {
     return std::make_shared<IndexItem>(object == Py_True);
   } else if (object == Py_None) {

oneflow/api/python/functional/python_arg.cpp

@@ -17,6 +17,7 @@ limitations under the License.
 #include "oneflow/api/python/functional/python_arg.h"
 #include "oneflow/api/python/functional/common.h"
 #include "oneflow/api/python/functional/indexing.h"
+#include "oneflow/extension/python/numpy.h"
 #include "oneflow/core/common/scalar.h"
 #include "oneflow/core/framework/dtype.h"
 #include "oneflow/core/framework/device.h"
@@ -198,21 +199,25 @@ Maybe<bool> PythonArg::TypeCheck(ValueType type) const {
     case kUINT32:
     case kINT64:
     case kUINT64:
-    case kBOOL: return PyLong_Check(object_);
+    case kBOOL: return PyLong_Check(object_) || numpy::PyArrayCheckLongScalar(object_);
     case kINT32_LIST:
     case kUINT32_LIST:
     case kINT64_LIST:
     case kUINT64_LIST:
     case kBOOL_LIST: return PyLongSequenceCheck(object_) || (size_ > 0 && PyLong_Check(object_));
     case kFLOAT:
-    case kDOUBLE: return PyFloat_Check(object_) || PyLong_Check(object_);
+    case kDOUBLE:
+      return PyFloat_Check(object_) || PyLong_Check(object_)
+             || numpy::PyArrayCheckFloatScalar(object_) || numpy::PyArrayCheckLongScalar(object_);
     case kFLOAT_LIST:
     case kDOUBLE_LIST:
       return PyFloatSquenceCheck(object_)
              || (size_ > 0 && (PyFloat_Check(object_) || PyLong_Check(object_)));
     case kSTRING: return PyStringCheck(object_);
     case kSTRING_LIST: return PyStringSequenceCheck(object_);
-    case kSCALAR: return PyScalarCheck(object_);
+    case kSCALAR:
+      return PyScalarCheck(object_) || numpy::PyArrayCheckLongScalar(object_)
+             || numpy::PyArrayCheckFloatScalar(object_);
     case kTENSOR:
     case kTENSOR_REF: return PyTensorCheck(object_);
     case kTENSOR_TUPLE: return PyTensorTupleCheck(object_) || PyTensorSequenceCheck(object_);

oneflow/api/python/utils/tensor_utils.h

@@ -31,7 +31,6 @@ limitations under the License.
 #include "oneflow/core/register/ofblob.h"
 #include "oneflow/core/common/blocking_then_busy.h"
 #include "oneflow/core/vm/virtual_machine.h"
-#include "oneflow/extension/python/numpy.h"
 #include "oneflow/core/common/foreign_lock_helper.h"
 
 namespace py = pybind11;

oneflow/extension/python/numpy.cpp

@@ -90,6 +90,18 @@ std::vector<size_t> OFStrideToNumpyStride(const StrideVector& fixed_vec, const D
   return result;
 }
 
+bool PyArrayCheckLongScalar(PyObject* obj) {
+  return PyArray_CheckScalar(obj) && PyDataType_ISINTEGER(PyArray_DescrFromScalar(obj));
+}
+
+bool PyArrayCheckFloatScalar(PyObject* obj) {
+  return PyArray_CheckScalar(obj) && PyDataType_ISFLOAT(PyArray_DescrFromScalar(obj));
+}
+
+bool PyArrayCheckBoolScalar(PyObject* obj) {
+  return PyArray_CheckScalar(obj) && PyDataType_ISBOOL(PyArray_DescrFromScalar(obj));
+}
+
 // Executing any numpy c api before _import_array() results in segfault
 // NOTE: this InitNumpyCAPI() works because of `PY_ARRAY_UNIQUE_SYMBOL`
 // defined in numpy_internal.h

oneflow/extension/python/numpy_internal.h

@@ -62,6 +62,12 @@ std::vector<size_t> OFShapeToNumpyShape(const DimVector& fixed_vec);
 
 std::vector<size_t> OFStrideToNumpyStride(const StrideVector& fixed_vec, const DataType data_type);
 
+bool PyArrayCheckLongScalar(PyObject* obj);
+
+bool PyArrayCheckFloatScalar(PyObject* obj);
+
+bool PyArrayCheckBoolScalar(PyObject* obj);
+
 Maybe<void> InitNumpyCAPI();
 
 }  // namespace numpy

python/oneflow/test/tensor/test_tensor_indexing.py

@@ -23,6 +23,67 @@
 import oneflow.unittest
 
 
+def _test_numpy_scalar_indexing(test_case, numpy_x, np_scalar):
+    x = flow.Tensor(numpy_x)
+
+    # basic_slice
+    test_case.assertTrue(np.allclose(numpy_x[np_scalar(1)], x[np_scalar(1)].numpy()))
+    test_case.assertTrue(np.allclose(numpy_x[np_scalar(-2)], x[np_scalar(-2)].numpy()))
+    test_case.assertTrue(
+        np.allclose(
+            numpy_x[np_scalar(0), np_scalar(1)], x[np_scalar(0), np_scalar(1)].numpy()
+        )
+    )
+    test_case.assertTrue(
+        np.allclose(
+            numpy_x[(np_scalar(0), np_scalar(1))],
+            x[(np_scalar(0), np_scalar(1))].numpy(),
+        )
+    )
+    test_case.assertTrue(
+        np.allclose(
+            numpy_x[((np_scalar(0), np_scalar(1)))],
+            x[((np_scalar(0), np_scalar(1)))].numpy(),
+        )
+    )
+
+
+def _test_numpy_scalar_advance_indexing(test_case, numpy_x, np_scalar):
+    x = flow.Tensor(numpy_x)
+
+    # advance indexing
+    test_case.assertTrue(
+        np.allclose(
+            numpy_x[[np_scalar(0), np_scalar(1)]],
+            x[[np_scalar(0), np_scalar(1)]].numpy(),
+        )
+    )
+    test_case.assertTrue(
+        np.allclose(
+            numpy_x[[np_scalar(0), np_scalar(1)], [np_scalar(1), np_scalar(0)]],
+            x[[np_scalar(0), np_scalar(1)], [np_scalar(1), np_scalar(0)]].numpy(),
+        )
+    )
+    test_case.assertTrue(
+        np.allclose(
+            numpy_x[
+                [
+                    [np_scalar(0), np_scalar(1)],
+                    [np_scalar(0), np_scalar(1)],
+                    [np_scalar(1), np_scalar(0)],
+                ]
+            ],
+            x[
+                [
+                    [np_scalar(0), np_scalar(1)],
+                    [np_scalar(0), np_scalar(1)],
+                    [np_scalar(1), np_scalar(0)],
+                ]
+            ].numpy(),
+        )
+    )
+
+
 def _test_basic_slice(test_case, numpy_x):
     x = flow.tensor(numpy_x)
 
@@ -278,6 +339,28 @@ def test_combining_indexing(test_case):
         numpy_x = np.arange(0, 720, 1).reshape([8, 9, 10]).astype(np.float32)
         _test_combining_indexing(test_case, numpy_x)
 
+    def test_numpy_scalar_indexing(test_case):
+        for np_scalar in [np.int8, np.int16, np.int32, np.int64]:
+            numpy_x = np.arange(0, 60, 1).reshape([3, 4, 5]).astype(np.float32)
+            _test_numpy_scalar_indexing(test_case, numpy_x, np_scalar)
+
+            numpy_x = np.arange(0, 360, 1).reshape([3, 4, 5, 6]).astype(np.float32)
+            _test_numpy_scalar_indexing(test_case, numpy_x, np_scalar)
+
+            numpy_x = np.arange(0, 720, 1).reshape([8, 9, 10]).astype(np.float32)
+            _test_numpy_scalar_indexing(test_case, numpy_x, np_scalar)
+
+        # TODO: add np.int16 when advance indexing supports np.int16 mapping
+        for np_scalar in [np.int32, np.int64]:
+            numpy_x = np.arange(0, 60, 1).reshape([3, 4, 5]).astype(np.float32)
+            _test_numpy_scalar_advance_indexing(test_case, numpy_x, np_scalar)
+
+            numpy_x = np.arange(0, 360, 1).reshape([3, 4, 5, 6]).astype(np.float32)
+            _test_numpy_scalar_advance_indexing(test_case, numpy_x, np_scalar)
+
+            numpy_x = np.arange(0, 720, 1).reshape([8, 9, 10]).astype(np.float32)
+            _test_numpy_scalar_advance_indexing(test_case, numpy_x, np_scalar)
+
     def test_mask_getitem(test_case):
         numpy_x = np.arange(0, 60, 1).reshape([3, 4, 5]).astype(np.float32)
         _test_mask_getitem(test_case, numpy_x)