import torch import torch.nn.functional as F import torchaudio from transformers import AutoConfig, Wav2Vec2Processor from Wav2Vec2ForSpeechClassification import Wav2Vec2ForSpeechClassification MY_MODEL = "padmalcom/wav2vec2-large-nonverbalvocalization-classification" device = torch.device("cuda" if torch.cuda.is_available() else "cpu") config = AutoConfig.from_pretrained(MY_MODEL) processor = Wav2Vec2Processor.from_pretrained(MY_MODEL) sampling_rate = processor.feature_extractor.sampling_rate model = Wav2Vec2ForSpeechClassification.from_pretrained(MY_MODEL).to(device) def speech_file_to_array_fn(path, sampling_rate): speech_array, _sampling_rate = torchaudio.load(path) resampler = torchaudio.transforms.Resample(_sampling_rate) speech = resampler(speech_array).squeeze().numpy() return speech def predict(path, sampling_rate): speech = speech_file_to_array_fn(path, sampling_rate) features = processor(speech, sampling_rate=sampling_rate, return_tensors="pt", padding=True) input_values = features.input_values.to(device) attention_mask = features.attention_mask.to(device) with torch.no_grad(): logits = model(input_values, attention_mask=attention_mask).logits scores = F.softmax(logits, dim=1).detach().cpu().numpy()[0] outputs = [{"Vocalization": config.id2label[i], "Score": f"{round(score * 100, 3):.1f}%"} for i, score in enumerate(scores)] return outputs res = predict("test.wav", 16000) max = max(res, key=lambda x: x['Score']) print("Expected lip popping:", max)