[XPU] add fused_softmax_mask and fused_softmax_mask_grad. (#55914)

paddle/phi/backends/xpu/xpu2_op_list.cc

@@ -724,8 +724,10 @@ XPUOpMap& get_kl2_ops() {
                      phi::DataType::INT64})},
       {"softmax",
        XPUKernelSet({phi::DataType::FLOAT32, phi::DataType::FLOAT16})},
+      {"fused_softmax_mask", XPUKernelSet({phi::DataType::FLOAT32})},
       {"softmax_grad",
        XPUKernelSet({phi::DataType::FLOAT32, phi::DataType::FLOAT16})},
+      {"fused_softmax_mask_grad", XPUKernelSet({phi::DataType::FLOAT32})},
       {"softmax_with_cross_entropy_grad",
        XPUKernelSet({phi::DataType::FLOAT32, phi::DataType::FLOAT16})},
       {"softmax_with_cross_entropy",

paddle/phi/kernels/fusion/xpu/fused_softmax_mask_grad_kernel.cc

+// Copyright (c) 2023 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/phi/core/kernel_registry.h"
+#include "paddle/phi/kernels/softmax_grad_kernel.h"
+
+namespace phi {
+namespace fusion {
+
+template <typename T, typename Context>
+void FusedSoftmaxMaskGradKernel(const Context& dev_ctx,
+                                const DenseTensor& out,
+                                const DenseTensor& out_grad,
+                                DenseTensor* x_grad) {
+  dev_ctx.template Alloc<T>(x_grad);
+  SoftmaxGradKernel<T, Context>(
+      dev_ctx, out, out_grad, 3, x_grad);  // axis for softmax
+}
+
+}  // namespace fusion
+}  // namespace phi
+
+PD_REGISTER_KERNEL(fused_softmax_mask_grad,
+                   XPU,
+                   ALL_LAYOUT,
+                   phi::fusion::FusedSoftmaxMaskGradKernel,
+                   float) {}

paddle/phi/kernels/fusion/xpu/fused_softmax_mask_kernel.cc

+// Copyright (c) 2023 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/phi/backends/xpu/enforce_xpu.h"
+#include "paddle/phi/core/kernel_registry.h"
+
+namespace phi {
+namespace fusion {
+
+template <typename T, typename Context>
+void FusedSoftmaxMaskKernel(const Context& dev_ctx,
+                            const DenseTensor& x,
+                            const DenseTensor& mask,
+                            DenseTensor* out) {
+  auto* x_data = x.data<T>();
+  auto* mask_data = mask.data<T>();
+  auto* y_data = dev_ctx.template Alloc<T>(out);
+
+  auto x_dim = x.dims();
+  auto mask_dim = mask.dims();
+  auto query_seq_len = x_dim[2];
+  auto key_seq_len = x_dim[3];
+
+  PADDLE_ENFORCE_GT(query_seq_len,
+                    1,
+                    phi::errors::InvalidArgument(
+                        "Input x's second last dim must be large than 1 but "
+                        "received the second last dimension of x is %d",
+                        query_seq_len));
+
+  PADDLE_ENFORCE_EQ(key_seq_len >= 32 && key_seq_len < 8192,
+                    true,
+                    phi::errors::InvalidArgument(
+                        "Input x's last dim must be between [32, 8192) "
+                        "received the last dimension of x is %d",
+                        key_seq_len));
+
+  PADDLE_ENFORCE_EQ(mask_dim[1],
+                    1,
+                    phi::errors::InvalidArgument(
+                        "Input mask's second dim must be 1 "
+                        "received the second dimension of mask is %d",
+                        mask_dim[1]));
+
+  // dim of x and mask must be equal
+  for (size_t idx = 0; idx < 4; ++idx) {
+    if (idx == 1) continue;
+    PADDLE_ENFORCE_EQ(
+        x_dim[idx],
+        mask_dim[idx],
+        phi::errors::InvalidArgument(
+            "Input x's %dth dim should be equal with input mask's %dth dim "
+            "but "
+            "received the %dth dimension of x and mask are not equal "
+            "the %dth dim of x is %d, while the %dth dim of mask is %d.",
+            idx,
+            idx,
+            idx,
+            idx,
+            x_dim[idx],
+            idx,
+            mask_dim[idx]));
+  }
+  std::vector<int64_t> x_shape = phi::vectorize<int64_t>(x.dims());
+  std::vector<int64_t> mask_shape = phi::vectorize<int64_t>(mask.dims());
+
+  // int softmax_with_mask(Context* ctx, const T* x, const T* mask, T* y, const
+  // std::vector<int64_t>& x_shape, const std::vector<int64_t>& mask_shape);
+  int r = xpu::softmax_with_mask(
+      dev_ctx.x_context(), x_data, mask_data, y_data, x_shape, mask_shape);
+  PADDLE_ENFORCE_XDNN_SUCCESS(r, "softmax_with_mask");
+}
+
+}  // namespace fusion
+}  // namespace phi
+
+PD_REGISTER_KERNEL(fused_softmax_mask,
+                   XPU,
+                   ALL_LAYOUT,
+                   phi::fusion::FusedSoftmaxMaskKernel,
+                   float) {}

test/xpu/test_fused_softmax_mask_op_xpu.py

+#   Copyright (c) 2023 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.
+
+import unittest
+
+import numpy as np
+from get_test_cover_info import (
+    XPUOpTestWrapper,
+    create_test_class,
+    get_xpu_op_support_types,
+)
+from op_test_xpu import XPUOpTest
+
+import paddle
+
+paddle.enable_static()
+
+
+def _get_softmax(x, mask):
+    masked_x = (x + mask).astype("float32")
+    max_value = np.max(masked_x, axis=-1, keepdims=True)
+    before_exp = masked_x - max_value
+    exp = np.exp(before_exp)
+    exp_sum = np.sum(exp, axis=-1, keepdims=True)
+    rst = exp / exp_sum
+    return rst
+
+
+class XPUTestFusedSoftmaxMaskOp(XPUOpTestWrapper):
+    """Test sigmoid_cross_entropy_with_logit_op with binary label"""
+
+    def __init__(self):
+        self.op_name = "fused_softmax_mask"
+        self.use_dynamic_create_class = False
+
+    class TestFusedSoftmaxMaskOp(XPUOpTest):
+        def setUp(self):
+            self.set_xpu()
+            self.op_type = "fused_softmax_mask"
+            self.init_dtype()
+            self.set_output()
+
+        def set_output(self):
+            x = np.random.random((1, 4, 4096, 4096)).astype("float32")
+            mask_input = np.random.random((1, 1, 4096, 4096)).astype("float32")
+            self.inputs = {'X': x, 'Mask': mask_input}
+            rst = _get_softmax(x, mask_input)
+            self.outputs = {'Out': rst}
+
+        def test_check_output(self):
+            self.check_output_with_place(self.place)
+
+        def test_check_grad(self):
+            self.check_grad_with_place(self.place, ['X'], 'Out')
+
+        def set_xpu(self):
+            self.__class__.use_xpu = True
+            self.place = paddle.XPUPlace(0)
+
+        def init_dtype(self):
+            self.dtype = self.in_type
+
+
+support_types = get_xpu_op_support_types('fused_softmax_mask')
+for stype in support_types:
+    create_test_class(globals(), XPUTestFusedSoftmaxMaskOp, stype)
+
+if __name__ == '__main__':
+    unittest.main()