commit files to HF hub

asr_whisper_pipeline.py

@@ -0,0 +1,15 @@
+from transformers import AutoProcessor, AutoModelForSpeechSeq2Seq
+import torch, torchaudio
+
+def transcribe(audio_path):
+    model_name = "openai/whisper-tiny"
+    processor = AutoProcessor.from_pretrained(model_name)
+    model = AutoModelForSpeechSeq2Seq.from_pretrained(model_name)
+    speech, sr = torchaudio.load(audio_path)
+    if sr != 16000:
+        resampler = torchaudio.transforms.Resample(sr, 16000)
+        speech = resampler(speech)
+    inputs = processor(speech.squeeze(), sampling_rate=16000, return_tensors="pt")
+    with torch.no_grad():
+        ids = model.generate(**inputs)
+    return processor.batch_decode(ids, skip_special_tokens=True)[0]