MTG/instrument.py

import json
import hashlib
import random



PATH = "/work/fast_data_yinghao/MTG/mtg-jamendo-dataset/data"
# Load your dataset

number_to_letter = {
    0: "A",
    1: "B",
    2: "C",
    3: "D"
}

def get_instrument(_="test"):
    data_samples = []
    for split in [_]:
        input_file_path = f'{PATH}/splits/split-0/autotagging_instrument-{split}.tsv'  # Replace with the actual path to your dataset

        with open(input_file_path, "r") as f:
            for idx, line in enumerate(f):
                if idx > 0:
                    tmp = line.strip().split("\t")
                    instruments = tmp[5:]
                    audio_path = tmp[3]
                    audioid = "66_MTG-Jamendo_" +  audio_path.split("/")[-1]
                    if "low" not in audioid:
                        audioid = audioid[:-4] + ".low.mp3"

                    data_sample = {
                        "instruction": "Please provide the instrument of given audio.",
                        "input": f"<|SOA|>f'{audio_path}'<|EOA|>",
                        "output": ", ".join(sorted([instrument.split("-")[-1] for instrument in instruments])),
                        "uuid": audio_path,
                        "audioid": audio_path,
                        "split": [split if split != "validation" else "dev"],
                        "task_type": {"major": ["global_MIR"], "minor": ["instrument_classification"]},
                        "domain": "music",
                        "source": "MTG",
                        "other": {"tag":"null"}
                    }
                    data_samples.append(data_sample)
                # if idx > 2:
                #     break
        f.close()
        

    existed_uuid_list = set()
    all_instruments = set(instrument for data_sample in data_samples for instrument in data_sample["output"].split(", "))
    for data_sample in data_samples:
        # change testset instruction to the choice format
        data_sample["instruction"] = data_sample["instruction"] + " If you can find multiple instruments, please output in alphabeta order. Use ', ' to split multiple tags."
                
        # change uuid
        uuid_string = f"{data_sample['instruction']}#{data_sample['input']}#{data_sample['output']}"
        unique_id = hashlib.md5(uuid_string.encode()).hexdigest()[:16] #只取前16位

        if unique_id in existed_uuid_list:
            sha1_hash = hashlib.sha1(uuid_string.encode()).hexdigest()[:16] # 为了相加的时候位数对应上 # 将 MD5 和 SHA1 结果相加,并计算新的 MD5 作为最终的 UUID
            unique_id = hashlib.md5((unique_id + sha1_hash).encode()).hexdigest()[:16]

        existed_uuid_list.add(unique_id)
        data_sample["uuid"] = f"{unique_id}"
        
    return data_samples

if __name__ == "__main__":
    print("start")
    for split in ["test", "train", "validation"]:
        data_samples = get_instrument(split)
        # Save to JSONL format
        output_file_path = f'instrument_{split}.jsonl'
        
        with open(output_file_path, 'w') as outfile:
            for sample in data_samples:
                json.dump(sample, outfile)

                outfile.write('\n')
        outfile.close()