描述 #

使用MossFormer2_SE_48K进行音频降噪

核心功能 #

• 使用MossFormer2_SE_48K进行音频降噪
• 支持本地文件、TOS存储(tos,s3,http,https等schema)

格式支持 #

• MP3 (.mp3)
• WAV (.wav)
• FLAC (.flac)
• OGG (.ogg)
• AAC (.aac)
• M4A (.m4a)

算子参数 #

输入 #

输出 #

降噪成功输出降噪后音频路径，失败为None

参数 #

如参数没有默认值，则为必填参数

调用示例 #

下面的代码展示了如何使用 daft 运行算子对音频进行降噪。

from __future__ import annotations

import logging
import os

import ray

import daft
from daft import col
from daft.las.functions.audio.audio_denoise import AudioDenoise
from daft.las.functions.udf import las_udf

def configure_logging():
    logging.basicConfig(
        level=logging.INFO,
        format="%(asctime)s - %(name)s - %(levelname)s - %(message)s",
        datefmt="%Y-%m-%d %H:%M:%S.%s".format(),
    )
    logging.getLogger("tracing.span").setLevel(logging.WARNING)
    logging.getLogger("daft_io.stats").setLevel(logging.WARNING)
    logging.getLogger("DaftStatisticsManager").setLevel(logging.WARNING)
    logging.getLogger("DaftFlotillaScheduler").setLevel(logging.WARNING)
    logging.getLogger("DaftFlotillaDispatcher").setLevel(logging.WARNING)

configure_logging()

if __name__ == "__main__":
    TOS_INPUT_DIR_URL = os.getenv("TOS_INPUT_DIR_URL", "las-cn-beijing-public-online.tos-cn-beijing.volces.com")
    TOS_OUTPUT_DIR = os.getenv("TOS_OUTPUT_DIR", "las-cn-beijing-public-online")
    samples = {
        "input_path": [os.path.join(f"https://{TOS_INPUT_DIR_URL}", "public/shared_audio_dataset/黑神话悟空对话.mp3")],
        "output_path": [os.path.join(f"tos://{TOS_OUTPUT_DIR}", "public/output/黑神话悟空对话_denoised.mp3")],
    }
    # 输出至output_path需要设置tos access_key和secret_key等认证信息,其中火山环境中TOS_ENDPOINT可使用内网地址以提升上传下载速度
    # os.environ["TOS_ACCESS_KEY"] = os.getenv("TOS_ACCESS_KEY", "aksk")
    # os.environ["TOS_SECRET_KEY"] = os.getenv("TOS_SECRET_KEY", "aksk")
    # os.environ["TOS_ENDPOINT"] = os.getenv("TOS_ENDPOINT", "https://tos-cn-beijing.volces.com")
    # os.environ["TOS_REGION"] = os.getenv("TOS_REGION", "cn-beijing")

    ray.init(dashboard_host="0.0.0.0", runtime_env={"worker_process_setup_hook": configure_logging})
    daft.context.set_runner_ray()

    daft.set_execution_config(actor_udf_ready_timeout=600)
    daft.set_execution_config(min_cpu_per_task=0)

    df_samples = daft.from_pydict(samples)

    df = df_samples.with_column(
        "result_path",
        las_udf(
            AudioDenoise,
            construct_args={
                "model_path": "/opt/las/models",
            },
            num_gpus=1,
            concurrency=1,
            batch_size=2,
        )(col("input_path"), col("output_path")),
    )

    df.show(max_width=120, format="grid")

    # ┌────────────────────────────────────┬───────────────────────────────────────────────────────────────────┬────────────────────────────────────────────────────────────────────┐
    # │ audio_path                         │ output_path                                                       │ result_path                                                        │
    # ╞════════════════════════════════════╪═══════════════════════════════════════════════════════════════════╪════════════════════════════════════════════════════════════════════╡
    # │ tos://xxxxx/黑神话悟空对话.mp3 │ tos://xxxxx/output/黑神话悟空对话_denoised.mp3 │ tos://xxxxx/output/黑神话悟空对话_denoised.mp3 │
    # └────────────────────────────────────┴───────────────────────────────────────────────────────────────────┴────────────────────────────────────────────────────────────────────┘