diff --git a/egs/librispeech/ASR/conv_emformer_transducer_stateless/train.py b/egs/librispeech/ASR/conv_emformer_transducer_stateless/train.py
index 106f3e511..6c3e01fe8 100755
--- a/egs/librispeech/ASR/conv_emformer_transducer_stateless/train.py
+++ b/egs/librispeech/ASR/conv_emformer_transducer_stateless/train.py
@@ -686,6 +686,14 @@ def compute_loss(
             (feature_lens // params.subsampling_factor).sum().item()
         )
 
+    info["utterances"] = feature.size(0)
+    # `utt_duration` and `utt_pad_proportion` would be normalized by `utterances`  # noqa
+    info["utt_duration"] = feature_lens.sum().item()
+    # padding proportion of each utterance
+    info["utt_pad_proportion"] = (
+        ((feature.size(1) - feature_lens) / feature.size(1)).sum().item()
+    )
+
     # Note: We use reduction=sum while computing the loss.
     info["loss"] = loss.detach().cpu().item()
     info["simple_loss"] = simple_loss.detach().cpu().item()
diff --git a/egs/librispeech/ASR/pruned_transducer_stateless/train.py b/egs/librispeech/ASR/pruned_transducer_stateless/train.py
index 448419759..626793815 100755
--- a/egs/librispeech/ASR/pruned_transducer_stateless/train.py
+++ b/egs/librispeech/ASR/pruned_transducer_stateless/train.py
@@ -504,6 +504,14 @@ def compute_loss(
             (feature_lens // params.subsampling_factor).sum().item()
         )
 
+    info["utterances"] = feature.size(0)
+    # `utt_duration` and `utt_pad_proportion` would be normalized by `utterances`  # noqa
+    info["utt_duration"] = feature_lens.sum().item()
+    # padding proportion of each utterance
+    info["utt_pad_proportion"] = (
+        ((feature.size(1) - feature_lens) / feature.size(1)).sum().item()
+    )
+
     # Note: We use reduction=sum while computing the loss.
     info["loss"] = loss.detach().cpu().item()
     info["simple_loss"] = simple_loss.detach().cpu().item()
diff --git a/egs/librispeech/ASR/pruned_transducer_stateless2/train.py b/egs/librispeech/ASR/pruned_transducer_stateless2/train.py
index 36ee7ca74..32e7cf5b4 100755
--- a/egs/librispeech/ASR/pruned_transducer_stateless2/train.py
+++ b/egs/librispeech/ASR/pruned_transducer_stateless2/train.py
@@ -573,6 +573,14 @@ def compute_loss(
             (feature_lens // params.subsampling_factor).sum().item()
         )
 
+    info["utterances"] = feature.size(0)
+    # `utt_duration` and `utt_pad_proportion` would be normalized by `utterances`  # noqa
+    info["utt_duration"] = feature_lens.sum().item()
+    # padding proportion of each utterance
+    info["utt_pad_proportion"] = (
+        ((feature.size(1) - feature_lens) / feature.size(1)).sum().item()
+    )
+
     # Note: We use reduction=sum while computing the loss.
     info["loss"] = loss.detach().cpu().item()
     info["simple_loss"] = simple_loss.detach().cpu().item()
diff --git a/egs/librispeech/ASR/pruned_transducer_stateless3/train.py b/egs/librispeech/ASR/pruned_transducer_stateless3/train.py
index 92eae78d1..4f7ccf111 100755
--- a/egs/librispeech/ASR/pruned_transducer_stateless3/train.py
+++ b/egs/librispeech/ASR/pruned_transducer_stateless3/train.py
@@ -612,6 +612,14 @@ def compute_loss(
             (feature_lens // params.subsampling_factor).sum().item()
         )
 
+    info["utterances"] = feature.size(0)
+    # `utt_duration` and `utt_pad_proportion` would be normalized by `utterances`  # noqa
+    info["utt_duration"] = feature_lens.sum().item()
+    # padding proportion of each utterance
+    info["utt_pad_proportion"] = (
+        ((feature.size(1) - feature_lens) / feature.size(1)).sum().item()
+    )
+
     # Note: We use reduction=sum while computing the loss.
     info["loss"] = loss.detach().cpu().item()
     info["simple_loss"] = simple_loss.detach().cpu().item()
diff --git a/egs/librispeech/ASR/pruned_transducer_stateless5/train.py b/egs/librispeech/ASR/pruned_transducer_stateless5/train.py
index e77eb19ff..03754f65a 100755
--- a/egs/librispeech/ASR/pruned_transducer_stateless5/train.py
+++ b/egs/librispeech/ASR/pruned_transducer_stateless5/train.py
@@ -644,6 +644,14 @@ def compute_loss(
             (feature_lens // params.subsampling_factor).sum().item()
         )
 
+    info["utterances"] = feature.size(0)
+    # `utt_duration` and `utt_pad_proportion` would be normalized by `utterances`  # noqa
+    info["utt_duration"] = feature_lens.sum().item()
+    # padding proportion of each utterance
+    info["utt_pad_proportion"] = (
+        ((feature.size(1) - feature_lens) / feature.size(1)).sum().item()
+    )
+
     # Note: We use reduction=sum while computing the loss.
     info["loss"] = loss.detach().cpu().item()
     info["simple_loss"] = simple_loss.detach().cpu().item()
diff --git a/egs/librispeech/ASR/pruned_transducer_stateless6/train.py b/egs/librispeech/ASR/pruned_transducer_stateless6/train.py
index 315c01c8e..136b20b6f 100755
--- a/egs/librispeech/ASR/pruned_transducer_stateless6/train.py
+++ b/egs/librispeech/ASR/pruned_transducer_stateless6/train.py
@@ -657,6 +657,14 @@ def compute_loss(
             (feature_lens // params.subsampling_factor).sum().item()
         )
 
+    info["utterances"] = feature.size(0)
+    # `utt_duration` and `utt_pad_proportion` would be normalized by `utterances`  # noqa
+    info["utt_duration"] = feature_lens.sum().item()
+    # padding proportion of each utterance
+    info["utt_pad_proportion"] = (
+        ((feature.size(1) - feature_lens) / feature.size(1)).sum().item()
+    )
+
     # Note: We use reduction=sum while computing the loss.
     info["loss"] = loss.detach().cpu().item()
     info["simple_loss"] = simple_loss.detach().cpu().item()