【Hackathon 5th No.52】 为 Paddle 新增 squeeze 和 unsqueeze 的 spmd 切分推导规则 (#57877)

WintersMontagne10335 · web-flow · commit 708acf00afea · 2023-11-27T16:09:08.000+08:00
* Add spmd segmentation and derivation rules for squeeze to Paddle

* Add spmd segmentation derivation rule for unsqueeze to Paddle

* fix bugs

* fix bugs

* fix bugs

* fix bugs

* Add unit test code

* modify squeeze.cc and CMakeLists.txt

* write separate rules

* fix bugs

* fix bugs

* fix bugs

* remove unsqueeze spmd rule

* modified:   test/auto_parallel/spmd_rules/test_squeeze_rule.py

* re-run CI

* fix bugs

* modify pointer to smart pointer

* fix bugs

* fix bugs
diff --git a/paddle/phi/infermeta/spmd_rules/rules.h b/paddle/phi/infermeta/spmd_rules/rules.h
@@ -33,6 +33,7 @@ limitations under the License. */
 #include "paddle/phi/infermeta/spmd_rules/slice.h"
 #include "paddle/phi/infermeta/spmd_rules/softmax.h"
 #include "paddle/phi/infermeta/spmd_rules/split.h"
+#include "paddle/phi/infermeta/spmd_rules/squeeze.h"
 #include "paddle/phi/infermeta/spmd_rules/stack.h"
 #include "paddle/phi/infermeta/spmd_rules/transpose.h"
 #include "paddle/phi/infermeta/spmd_rules/triu.h"
@@ -520,6 +521,10 @@ PD_REGISTER_SPMD_RULE(reshape2,
                       PD_INFER_SPMD(phi::distributed::ReshapeInferSpmd),
                       PD_INFER_SPMD(phi::distributed::ReshapeInferSpmdReverse));
 
+// squeeze rule
+PD_REGISTER_SPMD_RULE(squeeze,
+                      PD_INFER_SPMD(phi::distributed::SqueezeInferSpmd),
+                      PD_INFER_SPMD(phi::distributed::SqueezeInferSpmdReverse));
 // flatten rule
 PD_REGISTER_SPMD_RULE(flatten,
                       PD_INFER_SPMD(phi::distributed::FlattenInferSpmd),
diff --git a/paddle/phi/infermeta/spmd_rules/squeeze.cc b/paddle/phi/infermeta/spmd_rules/squeeze.cc
@@ -0,0 +1,222 @@
+/* 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
+
+https://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/infermeta/spmd_rules/squeeze.h"
+#include <algorithm>
+#include <numeric>
+
+#include "glog/logging.h"
+
+#include "paddle/phi/core/distributed/auto_parallel/dist_attr.h"
+#include "paddle/phi/core/distributed/auto_parallel/inferspmd_utils.h"
+#include "paddle/phi/core/distributed/auto_parallel/utils.h"
+#include "paddle/phi/infermeta/spmd_rules/dim_trans.h"
+#include "paddle/phi/infermeta/spmd_rules/utils.h"
+
+namespace phi {
+namespace distributed {
+
+using phi::distributed::auto_parallel::str_join;
+
+void MakeSqueezeDimTransWithoutAxis(
+    const std::vector<int64_t>& x_shape,
+    std::vector<int64_t>* out_shape,
+    std::vector<std::shared_ptr<DimTrans>>* trans) {
+  for (int64_t i = 0, n = static_cast<int64_t>(x_shape.size()); i < n; i++) {
+    if (x_shape[i] != 1) {
+      trans->emplace_back(std::make_shared<InputDim>(i));
+      out_shape->emplace_back(x_shape[i]);
+    }
+  }
+}
+
+void MakeSqueezeDimTransWithAxis(
+    const std::vector<int64_t>& x_shape,
+    std::vector<int64_t>* out_shape,
+    const std::vector<int64_t>& axis,
+    std::vector<std::shared_ptr<DimTrans>>* trans) {
+  for (int64_t i = 0, n = static_cast<int64_t>(x_shape.size()); i < n; i++) {
+    if (x_shape[i] == 1) {
+      auto it = find(axis.begin(), axis.end(), i);
+      if (it == axis.end()) {
+        trans->emplace_back(std::make_shared<Singleton>());
+        out_shape->emplace_back(1);
+      }
+    } else {
+      trans->emplace_back(std::make_shared<InputDim>(i));
+      out_shape->emplace_back(x_shape[i]);
+    }
+  }
+}
+
+void MakeSqueezeDimTransReverseWithoutAxis(
+    const std::vector<int64_t>& x_shape,
+    std::vector<std::shared_ptr<DimTrans>>* trans) {
+  for (int64_t i = 0, j = 0, n = static_cast<int64_t>(x_shape.size()); i < n;
+       i++) {
+    if (x_shape[i] != 1) {
+      trans->emplace_back(std::make_shared<InputDim>(j++));
+    } else {
+      trans->emplace_back(std::make_shared<Singleton>());
+    }
+  }
+}
+
+void MakeSqueezeDimTransReverseWithAxis(
+    const std::vector<int64_t>& x_shape,
+    const std::vector<int64_t>& out_shape,
+    const std::vector<int64_t>& axis,
+    std::vector<std::shared_ptr<DimTrans>>* trans) {
+  for (int64_t i = 0, j = 0, n = static_cast<int64_t>(x_shape.size()); i < n;
+       i++) {
+    if (x_shape[i] == 1) {
+      trans->emplace_back(std::make_shared<Singleton>());
+
+      auto it = find(axis.begin(), axis.end(), i);
+      if (it == axis.end()) {
+        j++;
+      }
+    } else {
+      trans->emplace_back(std::make_shared<InputDim>(j++));
+    }
+  }
+}
+
+SpmdInfo SqueezeInferSpmd(const DistMetaTensor& x,
+                          const std::vector<int64_t>& axis) {
+  // Step0: Verify input args based on squeeze logic
+  auto x_shape = phi::vectorize(x.dims());
+  int x_ndim = x_shape.size();
+  auto x_dist_attr_src = x.dist_attr();
+  std::vector<int64_t> x_dims_mapping = x_dist_attr_src.dims_mapping();
+
+  PADDLE_ENFORCE_EQ(
+      x_ndim,
+      x_dims_mapping.size(),
+      phi::errors::InvalidArgument("The Tensor X's rank [%d] and X's "
+                                   "dims_mapping size [%d] are not matched.",
+                                   x_ndim,
+                                   x_dims_mapping.size()));
+
+  // Step1: Build the transformation from
+  // the original shape to the target shape
+
+  std::vector<std::shared_ptr<DimTrans>> trans;
+  std::vector<int64_t> out_shape;
+
+  if (static_cast<int64_t>(axis.size()) == 0) {
+    MakeSqueezeDimTransWithoutAxis(x_shape, &out_shape, &trans);
+  } else {
+    std::vector<int64_t> axis_copy(axis);
+    for (int64_t i = 0, n = static_cast<int64_t>(axis_copy.size()); i < n;
+         i++) {
+      if (axis_copy[i] < 0) {
+        axis_copy[i] += x_ndim;
+      }
+    }
+    MakeSqueezeDimTransWithAxis(x_shape, &out_shape, axis_copy, &trans);
+  }
+
+  // Step2: Infer the dims mapping of input (if reshard is
+  // needed) and output from the dimension transformation.
+  std::vector<std::vector<int64_t>> dims_mapping_vec =
+      InferFromDimTrans(x, trans);
+
+  // Step3: Update the dist attributes of input
+  // and output with the inferred dims mapping.
+  TensorDistAttr x_dist_attr_dst(x_dist_attr_src);
+  x_dist_attr_dst.set_dims_mapping(dims_mapping_vec[0]);
+  TensorDistAttr out_dist_attr(x_dist_attr_src);
+  out_dist_attr.set_dims_mapping(dims_mapping_vec[1]);
+
+  VLOG(4) << "SqueezeInferSpmd: X shape: [" << str_join(x_shape)
+          << "] Out shape: [" << str_join(out_shape) << "]";
+  VLOG(4) << "Transformation from input to output:";
+  for (int64_t i = 0, n = static_cast<int64_t>(trans.size()); i < n; i++) {
+    VLOG(4) << "\tOut axis[" << i << "]: " << trans[i]->to_string();
+  }
+  VLOG(4) << "X dims_mapping_src: [" << str_join(x_dims_mapping)
+          << "] dims_mapping_dst: [" << str_join(dims_mapping_vec[0])
+          << "]\n Out dims_mapping: [" << str_join(dims_mapping_vec[1])
+          << "]\n\n";
+
+  return {{x_dist_attr_dst}, {out_dist_attr}};
+}
+
+SpmdInfo SqueezeInferSpmdReverse(const DistMetaTensor& x,
+                                 const DistMetaTensor& out,
+                                 const std::vector<int64_t>& axis) {
+  // Step0: Verify input args based on squeeze logic
+  auto x_shape = phi::vectorize(x.dims());
+  int x_ndim = x_shape.size();
+  auto out_shape = phi::vectorize(out.dims());
+  int out_ndim = out_shape.size();
+  auto out_dist_attr_src = out.dist_attr();
+  std::vector<int64_t> out_dims_mapping = out_dist_attr_src.dims_mapping();
+  PADDLE_ENFORCE_EQ(
+      out_ndim,
+      out_dims_mapping.size(),
+      phi::errors::InvalidArgument("The Tensor Out's rank [%d] and Out's "
+                                   "dims_mapping size [%d] are not matched.",
+                                   out_ndim,
+                                   out_dims_mapping.size()));
+
+  // Step1: Build the transformation from the output shape
+  // to original shape. This function infers the dims mapping
+  // from output to input, we first get the transformation
+  // from output to input so that we can infer the dims mapping
+  // with the map from output axes to input axes.
+
+  std::vector<std::shared_ptr<DimTrans>> trans;
+
+  if (static_cast<int64_t>(axis.size()) == 0) {
+    MakeSqueezeDimTransReverseWithoutAxis(x_shape, &trans);
+  } else {
+    std::vector<int64_t> axis_copy(axis);
+    for (int64_t i = 0, n = static_cast<int64_t>(axis_copy.size()); i < n;
+         i++) {
+      if (axis_copy[i] < 0) {
+        axis_copy[i] += x_ndim;
+      }
+    }
+    MakeSqueezeDimTransReverseWithAxis(x_shape, out_shape, axis_copy, &trans);
+  }
+
+  // Step2: Infer the dims mapping of input with
+  // output's dims_mapping and the transformation.
+  std::vector<std::vector<int64_t>> dims_mapping_vec =
+      InferFromDimTrans(out, trans);
+
+  // Step3: Update the dist attributes of input
+  // and output with the inferred dims mapping
+  TensorDistAttr out_dist_attr_dst(out_dist_attr_src);
+  out_dist_attr_dst.set_dims_mapping(dims_mapping_vec[0]);
+  TensorDistAttr x_dist_attr(x.dist_attr());
+  x_dist_attr.set_dims_mapping(dims_mapping_vec[1]);
+
+  VLOG(4) << "SqueezeInferSpmdReverse: Out shape: [" << str_join(out_shape)
+          << "] X shape: [" << str_join(x_shape) << "]";
+  VLOG(4) << "Transformation from output to input:";
+  for (int64_t i = 0, n = trans.size(); i < n; i++) {
+    VLOG(4) << "\tX axis[" << i << "]: " << trans[i]->to_string();
+  }
+  VLOG(4) << "Out dims_mapping_src: [" << str_join(out_dims_mapping) << "] "
+          << "dims_mapping_dst: [" << str_join(dims_mapping_vec[0]) << "]";
+  VLOG(4) << "X dims_mapping: [" << str_join(dims_mapping_vec[1]) << "]\n\n";
+
+  return {{x_dist_attr}, {out_dist_attr_dst}};
+}
+
+}  // namespace distributed
+}  // namespace phi
diff --git a/paddle/phi/infermeta/spmd_rules/squeeze.h b/paddle/phi/infermeta/spmd_rules/squeeze.h
@@ -0,0 +1,32 @@
+/* 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
+
+https://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. */
+
+#pragma once
+
+#include <vector>
+
+#include "paddle/phi/core/distributed/auto_parallel/dist_meta_tensor.h"
+#include "paddle/phi/core/distributed/type_defs.h"
+
+namespace phi {
+namespace distributed {
+
+SpmdInfo SqueezeInferSpmd(const DistMetaTensor& x,
+                          const std::vector<int64_t>& axis);
+
+SpmdInfo SqueezeInferSpmdReverse(const DistMetaTensor& x,
+                                 const DistMetaTensor& out,
+                                 const std::vector<int64_t>& axis);
+}  // namespace distributed
+}  // namespace phi
diff --git a/test/auto_parallel/spmd_rules/CMakeLists.txt b/test/auto_parallel/spmd_rules/CMakeLists.txt
@@ -18,6 +18,7 @@ if(WITH_DISTRIBUTE)
   py_test_modules(test_default_data_parallel_rule MODULES
                   test_default_data_parallel_rule)
   py_test_modules(test_layer_norm_rule MODULES test_layer_norm_rule)
+  py_test_modules(test_squeeze_rule MODULES test_squeeze_rule)
   py_test_modules(test_slice_rule MODULES test_slice_rule)
   py_test_modules(test_flatten_rule MODULES test_flatten_rule)
   py_test_modules(test_unsqueeze_rule MODULES test_unsqueeze_rule)
diff --git a/test/auto_parallel/spmd_rules/test_squeeze_rule.py b/test/auto_parallel/spmd_rules/test_squeeze_rule.py
