From 22be4fd731e8d767a82772c6ae3c01c7d28c6c0b Mon Sep 17 00:00:00 2001
From: rnyak <ronayak@hotmail.com>
Date: Tue, 8 Nov 2022 13:02:17 -0800
Subject: [PATCH] fix typo and metric error (#522)

---
 tests/torch/test_ranking_metrics.py      | 4 ++--
 transformers4rec/torch/ranking_metric.py | 8 +++++---
 2 files changed, 7 insertions(+), 5 deletions(-)

diff --git a/tests/torch/test_ranking_metrics.py b/tests/torch/test_ranking_metrics.py
index f59c47dc58..9e0a558ab2 100644
--- a/tests/torch/test_ranking_metrics.py
+++ b/tests/torch/test_ranking_metrics.py
@@ -17,7 +17,7 @@
 import pytest
 import torch
 
-from transformers4rec.torch.ranking_metric import MeanRecipricolRankAt
+from transformers4rec.torch.ranking_metric import MeanReciprocalRankAt
 
 tr = pytest.importorskip("transformers4rec.torch")
 
@@ -47,7 +47,7 @@ def test_score_with_transform_onehot(torch_ranking_metrics_inputs, metric):
 
 
 def test_mean_recipricol_rank():
-    metric = MeanRecipricolRankAt()
+    metric = MeanReciprocalRankAt()
     metric.top_ks = [1, 2, 3, 4]
     metric.labels_onehot = False
     result = metric(
diff --git a/transformers4rec/torch/ranking_metric.py b/transformers4rec/torch/ranking_metric.py
index 94bae9b98f..9c3dcf07f4 100644
--- a/transformers4rec/torch/ranking_metric.py
+++ b/transformers4rec/torch/ranking_metric.py
@@ -280,9 +280,9 @@ def _metric(
 
 
 @ranking_metrics_registry.register_with_multiple_names("mrr_at", "mrr")
-class MeanRecipricolRankAt(RankingMetric):
+class MeanReciprocalRankAt(RankingMetric):
     def __init__(self, top_ks=None, labels_onehot=False):
-        super(MeanRecipricolRankAt, self).__init__(top_ks=top_ks, labels_onehot=labels_onehot)
+        super(MeanReciprocalRankAt, self).__init__(top_ks=top_ks, labels_onehot=labels_onehot)
 
     def _metric(
         self, ks: torch.Tensor, scores: torch.Tensor, labels: torch.Tensor, log_base: int = 2
@@ -310,6 +310,8 @@ def _metric(
             device=scores.device, dtype=torch.float32
         )
         for index, k in enumerate(ks):
-            values, _ = (topk_labels[:, :k] / (torch.arange(k) + 1)).max(dim=1)
+            values, _ = (topk_labels[:, :k] / (torch.arange(k) + 1).to(device=scores.device)).max(
+                dim=1
+            )
             results[:, index] = values
         return results