import os from os.path import expanduser import shutil import torch from soundfile import LibsndfileError from datasets import load_dataset, DatasetDict, Audio from tokenizer_encodec import EncodecTokenizer direction = os.getenv("DIRECTION", "enA-jaA") sides = set(direction.split("-")) dataset_id = os.getenv("DATASET_ID", 0) num_proc = int(os.getenv("NUM_PROC", 1)) hf_org = os.getenv("HF_ORG", "asahi417") hf_dataset = f"seamless-align-{direction}" dataset = load_dataset(f"{hf_org}/{hf_dataset}", f"subset_{dataset_id}", split="train") tokenizer = EncodecTokenizer.from_pretrained() audio_loader = Audio() def error_file(example): for side in sides: try: audio_loader.decode_example(example[f"{side}.audio"]) except LibsndfileError: return False return True print(f"Num examples: {len(dataset)}") for s in sides: dataset = dataset.cast_column(f"{s}.audio", Audio(decode=False)) dataset = dataset.filter(error_file, num_proc=num_proc, desc="drop broken audio") for s in sides: dataset = dataset.cast_column(f"{s}.audio", Audio()) print(f"Num examples (after filtering): {len(dataset)}") def tokenize(example): for side in sides: wav = torch.as_tensor(example[f"{side}.audio"]["array"].reshape(1, 1, -1), dtype=torch.float32) example[f"{side}.audio.tokens"] = tokenizer.wav_to_tokens( wav=wav, sample_rate=example[f"{side}.audio"]["sampling_rate"] ).numpy().tolist()[0] return example dataset = dataset.map( function=tokenize, remove_columns=[f"{s}.audio" for s in sides] + [f"{s}.url" for s in sides] + [f"{s}.duration_start" for s in sides] + [f"{s}.duration_end" for s in sides], num_proc=num_proc, desc="tokenize dataset" ) DatasetDict({"train": dataset}).push_to_hub(f"{hf_org}/{hf_dataset}.tokenized", config_name=f"subset_{dataset_id}") # DatasetDict({"train": dataset}).push_to_hub(f"{hf_org}/{hf_dataset}.tokenized.encodec", config_name=f"subset_{dataset_id}") cache_dir = f"{expanduser('~')}/.cache/huggingface/datasets/{hf_org}___{hf_dataset}/subset_{dataset_id}" if os.path.exists(cache_dir): shutil.rmtree(cache_dir)