init

pipeline/kotoba_whisper.py

@@ -216,7 +216,7 @@ class KotobaWhisperPipeline(AutomaticSpeechRecognitionPipeline):
                         item["is_last"] = m == len(timelines) - 1 and n == len(labels) - 1 and item["is_last"]
                         yield item
                 else:
-                    if inputs.shape[0] > self.feature_extractor.n_samples:
+                    if audio_array.shape[0] > self.feature_extractor.n_samples:
                         processed = self.feature_extractor(
                             audio_array,
                             sampling_rate=self.feature_extractor.sampling_rate,

pipeline/push_pipeline.py

@@ -13,7 +13,14 @@ PIPELINE_REGISTRY.register_pipeline(
     pt_model=WhisperForConditionalGeneration,
     tf_model=TFWhisperForConditionalGeneration
 )
-pipe = pipeline(task="kotoba-whisper", model="kotoba-tech/kotoba-whisper-v2.0", chunk_length_s=15, batch_size=16)
+pipe = pipeline(task="kotoba-whisper", model="kotoba-tech/kotoba-whisper-v2.0", batch_size=16)
+# pprint(pipe("fukabori_trimmed_more_more.mp3"))
+pprint(pipe(
+    "fukabori_trimmed_more_more.mp3",
+    add_silence_end=0.5,
+    add_silence_start=0.5,
+    chunk_length_s=15,
+))
 pipe.push_to_hub(model_alias)