add hour splits

libriheavy.py

@@ -169,11 +169,41 @@ class Libriheavy(datasets.GeneratorBasedBuilder):
                 name="odd",
                 gen_kwargs={"speaker_chunks": odd_speaker_chunks, "split": "odd"}
             ),
+            datasets.SplitGenerator(
+                name="even-100",
+                gen_kwargs={"speaker_chunks": even_speaker_chunks, "split": "even", "hours": 100}
+            ),
+            datasets.SplitGenerator(
+                name="odd-100",
+                gen_kwargs={"speaker_chunks": odd_speaker_chunks, "split": "odd", "hours": 100}
+            ),
+            datasets.SplitGenerator(
+                name="even-500",
+                gen_kwargs={"speaker_chunks": even_speaker_chunks, "split": "even", "hours": 500}
+            ),
+            datasets.SplitGenerator(
+                name="odd-500",
+                gen_kwargs={"speaker_chunks": odd_speaker_chunks, "split": "odd", "hours": 500}
+            ),
+            datasets.SplitGenerator(
+                name="even-1000",
+                gen_kwargs={"speaker_chunks": even_speaker_chunks, "split": "even", "hours": 1000}
+            ),
+            datasets.SplitGenerator(
+                name="odd-1000",
+                gen_kwargs={"speaker_chunks": odd_speaker_chunks, "split": "odd", "hours": 1000}
+            ),
         ]
 
-    def _generate_examples(self, speaker_chunks, split):
+    def _generate_examples(self, speaker_chunks, split, hours=None):
         """Yields examples."""
+        hours_streamed = 0
+        finish_stream = False
+        if hours is None:
+            hours = float("inf")
         for chunk in speaker_chunks:
+            if finish_stream:
+                break
             npz = dict(np.load(chunk["audio"], allow_pickle=True))
             utterances = npz.keys()
             with gzip.open(chunk["text"], "rt") as f:
@@ -211,7 +241,11 @@ class Libriheavy(datasets.GeneratorBasedBuilder):
                             "srmr": npz_item["overall_srmr"],
                         },
                     }
+                    hours_streamed += (utterance["word_segments"][-1][1] - utterance["word_segments"][0][0]) / 3600
                     yield chunk["speaker_id"] + "_" + utterance_id, result
+                    if hours_streamed >= hours:
+                        finish_stream = True
+                        break
             else:
                 # only use the last utterance
                 utterance_id = sorted(list(text.keys()))[-1]
@@ -244,4 +278,8 @@ class Libriheavy(datasets.GeneratorBasedBuilder):
                         "srmr": npz_item["overall_srmr"],
                     },
                 }
-                yield chunk["speaker_id"] + "_" + utterance_id, result
+                hours_streamed += (utterance["word_segments"][-1][1] - utterance["word_segments"][0][0]) / 3600
+                yield chunk["speaker_id"] + "_" + utterance_id, result
+                if hours_streamed >= hours:
+                    finish_stream = True
+                    break