Fix to #38693 (minimal UT) (#41026)

* Add UT

- Added missed data_layout

- Added missing conversions

- NDHWC added

- NDHWC support in data_transform

- another fix

- condddate change

- fix

u- fix

- fix

- fix

- fix

- fix

- fix to hack

- compilation fix

- fix to automatic merge

* - reduced UT

* - fix

* - lint

* - fix to lint

paddle/fluid/framework/data_layout_transform.h

@@ -59,6 +59,10 @@ inline MKLDNNMemoryFormat ToMKLDNNFormat(const DataLayout& layout) {
       return MKLDNNMemoryFormat::nhwc;
     case DataLayout::kNCHW:
       return MKLDNNMemoryFormat::nchw;
+    case DataLayout::kNCDHW:
+      return MKLDNNMemoryFormat::ncdhw;
+    case DataLayout::kNDHWC:
+      return MKLDNNMemoryFormat::ndhwc;
     default:
       PADDLE_THROW(platform::errors::InvalidArgument(
           "Fail to convert layout %s to MKLDNN format.",
@@ -72,6 +76,10 @@ inline DataLayout ToPaddleLayout(const MKLDNNMemoryFormat& format) {
       return DataLayout::kNHWC;
     case MKLDNNMemoryFormat::nchw:
       return DataLayout::kNCHW;
+    case MKLDNNMemoryFormat::ncdhw:
+      return DataLayout::kNCDHW;
+    case MKLDNNMemoryFormat::ndhwc:
+      return DataLayout::kNDHWC;
     default:
       PADDLE_THROW(platform::errors::InvalidArgument(
           "Fail to convert MKLDNN format to paddle layout."));

paddle/fluid/framework/data_transform.cc

@@ -63,7 +63,7 @@ void TransformData(const OpKernelType &expected_kernel_type,
         out.ShareDataWith(input_tensor);
         // For NHWC data we need reshape of tensors as MKL-DNN
         // is expecting NHWC dims description order
-        if (lin == DataLayout::kNHWC) {
+        if (lin == DataLayout::kNHWC || lin == DataLayout::kNDHWC) {
           platform::MatchShapeToLayout(&out, lin, lout);
           // We register only NHWC assuming that model is consistent e.g. either
           // NHWC or NCHW

paddle/fluid/inference/api/details/zero_copy_tensor.cc

@@ -579,7 +579,8 @@ std::vector<int> Tensor::shape() const {
     // be done. Similarly for dim==1 when you have just one possible
     // combination.
     if (tensor->dims().size() < 3) return phi::vectorize<int>(tensor->dims());
-    if (out_layout == paddle::framework::DataLayout::kNHWC) {
+    if (out_layout == paddle::framework::DataLayout::kNHWC ||
+        out_layout == paddle::framework::DataLayout::kNDHWC) {
       auto dims = phi::vectorize<int>(tensor->dims());
       std::rotate(dims.begin() + 1, dims.begin() + 2, dims.end());
       return dims;

paddle/fluid/platform/mkldnn_helper.h

@@ -20,6 +20,7 @@ limitations under the License. */
 #include <string>
 #include <utility>
 #include <vector>
+
 #include "dnnl.hpp"
 #include "paddle/fluid/framework/operator.h"
 #include "paddle/fluid/platform/place.h"
@@ -102,20 +103,22 @@ inline void MatchShapeToLayout(framework::Tensor* tensor_in,
 
   switch (from) {
     case framework::DataLayout::kMKLDNN:
-      if (to == framework::DataLayout::kNHWC) {
+      if ((to == framework::DataLayout::kNHWC) ||
+          (to == framework::DataLayout::kNDHWC)) {
         auto dims = phi::vectorize<int>(tensor_in->dims());
         std::rotate(dims.begin() + 1, dims.begin() + 2, dims.end());
         tensor_in->Resize(phi::make_ddim(dims));
-        VLOG(3) << "Rotating Shape from: kMKLDNN to: kNHWC output_shape"
+        VLOG(3) << "Rotating Shape from: kMKLDNN to: kNHWC/kNDHWC output_shape"
                 << print_dims(dims);
       }
       break;
     case framework::DataLayout::kNHWC:
+    case framework::DataLayout::kNDHWC:
       if (to == framework::DataLayout::kMKLDNN) {
         auto dims = phi::vectorize<int>(tensor_in->dims());
         std::rotate(dims.begin() + 1, dims.end() - 1, dims.end());
         tensor_in->Resize(phi::make_ddim(dims));
-        VLOG(3) << "Rotating Shape from: kNHWC to: kMKLDNN output_shape"
+        VLOG(3) << "Rotating Shape from: kNHWC/kNDHWC to: kMKLDNN output_shape"
                 << print_dims(dims);
       }
       break;
@@ -279,7 +282,12 @@ inline dnnl::memory::format_tag GetMKLDNNFormat(dnnl::memory::desc mem_desc) {
         return dnnl::memory::format_tag::acdeb;
       }
     } else if (inner_nblks == 1) {
-      if (inner_blks[0] == 8 && inner_idxs[0] == 0) {
+      if (inner_blks[0] == 4 && inner_idxs[0] == 1) {
+        if (strides[0] >= strides[1] && strides[1] >= strides[2] &&
+            strides[2] >= strides[3] && strides[3] >= strides[4]) {
+          return dnnl::memory::format_tag::aBcde4b;
+        }
+      } else if (inner_blks[0] == 8 && inner_idxs[0] == 0) {
         if (strides[0] >= strides[2] && strides[2] >= strides[3] &&
             strides[3] >= strides[4] && strides[4] >= strides[1]) {
           return dnnl::memory::format_tag::Acdeb8a;

paddle/phi/common/layout.h

@@ -26,13 +26,13 @@ enum class DataLayout {
   ANY = UNDEFINED,
   NHWC,
   NCHW,
+  NCDHW,
+  NDHWC,
   MKLDNN,
   SPARSE_COO,
   SPARSE_CSR,
   PSTRING_UNION,
   NUM_DATA_LAYOUTS,
-  NDHWC,
-  NCDHW,
   // See Note [ Why we need ALL in basic kernel key member? ]
   ALL_LAYOUT = UNDEFINED,
   // Note: Unify phi DataLayout and fluid::framework::DataLayout,

python/paddle/fluid/tests/unittests/ir/inference/CMakeLists.txt

@@ -128,6 +128,7 @@ if (WITH_MKLDNN)
   set_tests_properties(test_mkldnn_depthwise_conv_pass PROPERTIES TIMEOUT 120)
   set_tests_properties(test_mkldnn_reshape_transpose_matmul_fuse_pass PROPERTIES TIMEOUT 100)
   set_tests_properties(test_mkldnn_mish_op PROPERTIES TIMEOUT 300)
+  set_tests_properties(test_mkldnn_conv3d_op PROPERTIES TIMEOUT 300)
   set_tests_properties(test_mkldnn_prelu_op PROPERTIES TIMEOUT 300)
   set_tests_properties(test_conv_act_mkldnn_fuse_pass PROPERTIES TIMEOUT 120)
   set_tests_properties(test_conv_transpose_eltwiseadd_bn_fuse_pass PROPERTIES TIMEOUT 250)

python/paddle/fluid/tests/unittests/ir/inference/test_mkldnn_conv3d_op.py

+# Copyright (c) 2021 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.
+
+from auto_scan_test import MkldnnAutoScanTest, SkipReasons
+from program_config import TensorConfig, ProgramConfig, OpConfig
+import numpy as np
+import paddle.inference as paddle_infer
+from functools import partial
+from typing import Optional, List, Callable, Dict, Any, Set
+import unittest
+
+import hypothesis
+from hypothesis import given, settings, seed, example, assume
+import hypothesis.strategies as st
+
+
+class TestMkldnnConv3dOp(MkldnnAutoScanTest):
+    def is_program_valid(self, program_config: ProgramConfig) -> bool:
+        return True
+
+    def sample_program_configs(self, *args, **kwargs):
+        def generate_input(*args, **kwargs):
+            if kwargs["data_format"] == "NCDHW":
+                return np.random.random(
+                    [kwargs["batch_size"], 48, 64, 32, 64]).astype(np.float32)
+            else:
+                return np.random.random(
+                    [kwargs["batch_size"], 64, 32, 64, 48]).astype(np.float32)
+
+        def generate_weight(*args, **kwargs):
+            return np.random.random(
+                [16, int(48 / kwargs["groups"]), 3, 3, 3]).astype(np.float32)
+
+        conv3d_op = OpConfig(
+            type="conv3d",
+            inputs={"Input": ["input_data"],
+                    "Filter": ["conv_weight"]},
+            outputs={"Output": ["conv_output"]},
+            attrs={
+                "data_format": kwargs["data_format"],
+                "dilations": kwargs["dilations"],
+                "padding_algorithm": kwargs["padding_algorithm"],
+                "groups": kwargs["groups"],
+                "paddings": kwargs["paddings"],
+                "strides": kwargs["strides"],
+                "is_test": True
+            })
+
+        program_config = ProgramConfig(
+            ops=[conv3d_op],
+            weights={
+                "conv_weight":
+                TensorConfig(data_gen=partial(generate_weight, *args, **kwargs))
+            },
+            inputs={
+                "input_data":
+                TensorConfig(data_gen=partial(generate_input, *args, **kwargs))
+            },
+            outputs=["conv_output"])
+
+        yield program_config
+
+    def sample_predictor_configs(self, program_config):
+        config = self.create_inference_config(use_mkldnn=True)
+        yield config, (1e-5, 1e-5)
+
+    @given(
+        data_format=st.sampled_from(["NCDHW", "NDHWC"]),
+        dilations=st.sampled_from([[1, 2, 1]]),
+        padding_algorithm=st.sampled_from(["EXPLICIT"]),
+        groups=st.sampled_from([2]),
+        paddings=st.sampled_from([[0, 3, 2]]),
+        strides=st.sampled_from([[1, 2, 1]]),
+        batch_size=st.integers(
+            min_value=1, max_value=4), )
+    def test(self, *args, **kwargs):
+        self.run_test(*args, **kwargs)
+
+
+if __name__ == "__main__":
+    unittest.main()