{ "_name_or_path": "facebook/wav2vec2-base", "architectures": [ "Wav2Vec2ForSequenceClassification" ], "hidden_size": 768, "model_type": "wav2vec2", "num_hidden_layers": 12, "vocab_size": 32, "id2label": { "0": "Car_Horn", "1": "Children_Playing", "2": "Dog_Barking", "3": "Machinery_Hum", "4": "Siren_Emergency", "5": "Train_Whistle", "6": "Tire_Screech", "7": "Glass_Shattering" }, "label2id": { "Car_Horn": 0, "Children_Playing": 1, "Dog_Barking": 2, "Machinery_Hum": 3, "Siren_Emergency": 4, "Train_Whistle": 5, "Tire_Screech": 6, "Glass_Shattering": 7 }, "num_labels": 8, "problem_type": "single_label_classification", "sampling_rate": 16000, "transformers_version": "4.36.0" }