[YDF] Move loss options definition

Old: learner/gradient_boosted_trees/gradient_boosted_trees.proto
New: model/gradient_boosted_trees/gradient_boosted_trees.proto
PiperOrigin-RevId: 676396458

CHANGELOG.md

@@ -12,6 +12,12 @@ Changelog under `yggdrasil_decision_forests/port/python/CHANGELOG.md`.
     but can still be used.
 -   Allow configuring the truncation of NDCG losses.
 
+### Misc
+
+-   Loss options are now defined 
+    model/gradient_boosted_trees/gradient_boosted_trees.proto (previously
+    learner/gradient_boosted_trees/gradient_boosted_trees.proto)
+
 ## 1.10.0 - 2024-08-21
 
 ### Features

yggdrasil_decision_forests/learner/gradient_boosted_trees/gradient_boosted_trees.proto

@@ -177,66 +177,9 @@ message GradientBoostedTreesTrainingConfig {
   optional int32 early_stopping_initial_iteration = 37 [default = 10];
 
   oneof loss_options {
-    LambdaMartNdcg lambda_mart_ndcg = 12;
-    XeNdcg xe_ndcg = 26;
-    BinaryFocalLossOptions binary_focal_loss_options = 36;
-  }
-
-  message LambdaMartNdcg {
-    // If false, the gradient is computed using NDCG i.e. normalized-DCG. If
-    // false, the gradient is computed using DCG.
-    optional bool gradient_use_non_normalized_dcg = 1 [default = false];
-
-    // Number of candidates considered when computing the NDCG loss.
-    //
-    // NDCG losses are usually truncated at a particular rank level (generally
-    // between 4 and 10), i.e. only the highly ranked documents are considered
-    // when computing the rank. A smaller values results in a model with
-    // increased emphasis on the first results of the ranking.
-    //
-    // Note that the NDCG truncation of the cross-entropy NDCG loss must be
-    // configured separately.
-    optional int32 ndcg_truncation = 2 [default = 5];
-  }
-
-  message XeNdcg {
-    enum Gamma {
-      // For the time being, defaults to UNIFORM.
-      AUTO = 0;
-      // Gammas are sampled from a uniform distribution on [0, 1].
-      UNIFORM = 1;
-      // Gammas are set to 1 across the board. This is more appropriate for
-      // click datasets with a large number of documents per query.
-      ONE = 2;
-    }
-    optional Gamma gamma = 1 [default = UNIFORM];
-
-    // Number of candidates considered when computing the NDCG loss.
-    //
-    // NDCG losses are usually truncated at a particular rank level (generally
-    // between 4 and 10), i.e. only the highly ranked documents are considered
-    // when computing the rank. A smaller values results in a model with
-    // increased emphasis on the first results of the ranking.
-    //
-    // Note that the NDCG truncation of the classic NDCG loss must be configured
-    // separately.
-    optional int32 ndcg_truncation = 2 [default = 5];
-  }
-
-  message BinaryFocalLossOptions {
-    // Exponent of the misprediction multiplier in focal loss.
-    // Corresponds to the gamma parameter in
-    // https://arxiv.org/pdf/1708.02002.pdf
-    optional float misprediction_exponent = 1 [default = 2.0];
-
-    // A hypertuning coefficient to multiply the loss and its gradient(s) in
-    // case of a positive sample.
-    // Loss and gradient on positive samples will be multiplied by
-    // positive_sample_coefficient, on negative samples will be multiplied
-    // by (1 - positive_sample_coefficient)
-    // Corresponds to the 'alpha' parameter in
-    // https://arxiv.org/pdf/1708.02002.pdf
-    optional float positive_sample_coefficient = 2 [default = 0.5];
+    LossConfiguration.LambdaMartNdcg lambda_mart_ndcg = 12;
+    LossConfiguration.XeNdcg xe_ndcg = 26;
+    LossConfiguration.BinaryFocalLossOptions binary_focal_loss_options = 36;
   }
 
   // L2 regularization on the tree predictions i.e. on the value of the leaf.

yggdrasil_decision_forests/learner/gradient_boosted_trees/loss/BUILD

@@ -152,6 +152,7 @@ cc_library_ydf(
         "//yggdrasil_decision_forests/learner/decision_tree:training",
         "//yggdrasil_decision_forests/learner/gradient_boosted_trees:gradient_boosted_trees_cc_proto",
         "//yggdrasil_decision_forests/model:abstract_model_cc_proto",
+        "//yggdrasil_decision_forests/model/gradient_boosted_trees:gradient_boosted_trees_cc_proto",
         "//yggdrasil_decision_forests/utils:compatibility",
         "//yggdrasil_decision_forests/utils:concurrency",
         "//yggdrasil_decision_forests/utils:random",

yggdrasil_decision_forests/learner/gradient_boosted_trees/loss/loss_imp_cross_entropy_ndcg.cc

@@ -34,6 +34,7 @@
 #include "yggdrasil_decision_forests/learner/gradient_boosted_trees/loss/loss_interface.h"
 #include "yggdrasil_decision_forests/learner/gradient_boosted_trees/loss/loss_utils.h"
 #include "yggdrasil_decision_forests/model/abstract_model.pb.h"
+#include "yggdrasil_decision_forests/model/gradient_boosted_trees/gradient_boosted_trees.pb.h"
 #include "yggdrasil_decision_forests/utils/compatibility.h"
 #include "yggdrasil_decision_forests/utils/concurrency.h"
 #include "yggdrasil_decision_forests/utils/random.h"
@@ -104,11 +105,11 @@ absl::Status CrossEntropyNDCGLoss::UpdateGradients(
     params.resize(group_size);
 
     switch (gbt_config_.xe_ndcg().gamma()) {
-      case proto::GradientBoostedTreesTrainingConfig::XeNdcg::ONE:
+      case proto::LossConfiguration::XeNdcg::ONE:
         std::fill(params.begin(), params.end(), 1.f);
         break;
-      case proto::GradientBoostedTreesTrainingConfig::XeNdcg::AUTO:
-      case proto::GradientBoostedTreesTrainingConfig::XeNdcg::UNIFORM:
+      case proto::LossConfiguration::XeNdcg::AUTO:
+      case proto::LossConfiguration::XeNdcg::UNIFORM:
         for (int item_idx = 0; item_idx < group_size; item_idx++) {
           params[item_idx] = distribution(*random);
         }

yggdrasil_decision_forests/model/gradient_boosted_trees/gradient_boosted_trees.proto

@@ -127,3 +127,62 @@ extend model.proto.SerializedModel {
   optional GradientBoostedTreesSerializedModel
       gradient_boosted_trees_serialized_model = 1001;
 }
+
+message LossConfiguration {
+  message LambdaMartNdcg {
+    // If false, the gradient is computed using NDCG i.e. normalized-DCG. If
+    // false, the gradient is computed using DCG.
+    optional bool gradient_use_non_normalized_dcg = 1 [default = false];
+
+    // Number of candidates considered when computing the NDCG loss.
+    //
+    // NDCG losses are usually truncated at a particular rank level (generally
+    // between 4 and 10), i.e. only the highly ranked documents are considered
+    // when computing the rank. A smaller values results in a model with
+    // increased emphasis on the first results of the ranking.
+    //
+    // Note that the NDCG truncation of the cross-entropy NDCG loss must be
+    // configured separately.
+    optional int32 ndcg_truncation = 2 [default = 5];
+  }
+
+  message XeNdcg {
+    enum Gamma {
+      // For the time being, defaults to UNIFORM.
+      AUTO = 0;
+      // Gammas are sampled from a uniform distribution on [0, 1].
+      UNIFORM = 1;
+      // Gammas are set to 1 across the board. This is more appropriate for
+      // click datasets with a large number of documents per query.
+      ONE = 2;
+    }
+    optional Gamma gamma = 1 [default = UNIFORM];
+
+    // Number of candidates considered when computing the NDCG loss.
+    //
+    // NDCG losses are usually truncated at a particular rank level (generally
+    // between 4 and 10), i.e. only the highly ranked documents are considered
+    // when computing the rank. A smaller values results in a model with
+    // increased emphasis on the first results of the ranking.
+    //
+    // Note that the NDCG truncation of the classic NDCG loss must be configured
+    // separately.
+    optional int32 ndcg_truncation = 2 [default = 5];
+  }
+
+  message BinaryFocalLossOptions {
+    // Exponent of the misprediction multiplier in focal loss.
+    // Corresponds to the gamma parameter in
+    // https://arxiv.org/pdf/1708.02002.pdf
+    optional float misprediction_exponent = 1 [default = 2.0];
+
+    // A hypertuning coefficient to multiply the loss and its gradient(s) in
+    // case of a positive sample.
+    // Loss and gradient on positive samples will be multiplied by
+    // positive_sample_coefficient, on negative samples will be multiplied
+    // by (1 - positive_sample_coefficient)
+    // Corresponds to the 'alpha' parameter in
+    // https://arxiv.org/pdf/1708.02002.pdf
+    optional float positive_sample_coefficient = 2 [default = 0.5];
+  }
+}