goathead777/Zero_Shot_Inference

1import os2import glob3import sys4import argparse5import logging6import json7import subprocess8import traceback9 10import librosa11import numpy as np12from scipy.io.wavfile import read13import torch14import logging15 16logging.getLogger("numba").setLevel(logging.ERROR)17logging.getLogger("matplotlib").setLevel(logging.ERROR)18 19MATPLOTLIB_FLAG = False20 21# logging.basicConfig(stream=sys.stdout, level=logging.DEBUG)22logger = logging23 24 25def load_checkpoint(checkpoint_path, model, optimizer=None, skip_optimizer=False):26    assert os.path.isfile(checkpoint_path)27    checkpoint_dict = torch.load(checkpoint_path, map_location="cpu")28    iteration = checkpoint_dict["iteration"]29    learning_rate = checkpoint_dict["learning_rate"]30    if (31        optimizer is not None32        and not skip_optimizer33        and checkpoint_dict["optimizer"] is not None34    ):35        optimizer.load_state_dict(checkpoint_dict["optimizer"])36    saved_state_dict = checkpoint_dict["model"]37    if hasattr(model, "module"):38        state_dict = model.module.state_dict()39    else:40        state_dict = model.state_dict()41    new_state_dict = {}42    for k, v in state_dict.items():43        try:44            # assert "quantizer" not in k45            # print("load", k)46            new_state_dict[k] = saved_state_dict[k]47            assert saved_state_dict[k].shape == v.shape, (48                saved_state_dict[k].shape,49                v.shape,50            )51        except:52            traceback.print_exc()53            print(54                "error, %s is not in the checkpoint" % k55            )  # shape不对也会，比如text_embedding当cleaner修改时56            new_state_dict[k] = v57    if hasattr(model, "module"):58        model.module.load_state_dict(new_state_dict)59    else:60        model.load_state_dict(new_state_dict)61    print("load ")62    logger.info(63        "Loaded checkpoint '{}' (iteration {})".format(checkpoint_path, iteration)64    )65    return model, optimizer, learning_rate, iteration66 67 68def save_checkpoint(model, optimizer, learning_rate, iteration, checkpoint_path):69    logger.info(70        "Saving model and optimizer state at iteration {} to {}".format(71            iteration, checkpoint_path72        )73    )74    if hasattr(model, "module"):75        state_dict = model.module.state_dict()76    else:77        state_dict = model.state_dict()78    torch.save(79        {80            "model": state_dict,81            "iteration": iteration,82            "optimizer": optimizer.state_dict(),83            "learning_rate": learning_rate,84        },85        checkpoint_path,86    )87 88 89def summarize(90    writer,91    global_step,92    scalars={},93    histograms={},94    images={},95    audios={},96    audio_sampling_rate=22050,97):98    for k, v in scalars.items():99        writer.add_scalar(k, v, global_step)100    for k, v in histograms.items():101        writer.add_histogram(k, v, global_step)102    for k, v in images.items():103        writer.add_image(k, v, global_step, dataformats="HWC")104    for k, v in audios.items():105        writer.add_audio(k, v, global_step, audio_sampling_rate)106 107 108def latest_checkpoint_path(dir_path, regex="G_*.pth"):109    f_list = glob.glob(os.path.join(dir_path, regex))110    f_list.sort(key=lambda f: int("".join(filter(str.isdigit, f))))111    x = f_list[-1]112    print(x)113    return x114 115 116def plot_spectrogram_to_numpy(spectrogram):117    global MATPLOTLIB_FLAG118    if not MATPLOTLIB_FLAG:119        import matplotlib120 121        matplotlib.use("Agg")122        MATPLOTLIB_FLAG = True123        mpl_logger = logging.getLogger("matplotlib")124        mpl_logger.setLevel(logging.WARNING)125    import matplotlib.pylab as plt126    import numpy as np127 128    fig, ax = plt.subplots(figsize=(10, 2))129    im = ax.imshow(spectrogram, aspect="auto", origin="lower", interpolation="none")130    plt.colorbar(im, ax=ax)131    plt.xlabel("Frames")132    plt.ylabel("Channels")133    plt.tight_layout()134 135    fig.canvas.draw()136    data = np.fromstring(fig.canvas.tostring_rgb(), dtype=np.uint8, sep="")137    data = data.reshape(fig.canvas.get_width_height()[::-1] + (3,))138    plt.close()139    return data140 141 142def plot_alignment_to_numpy(alignment, info=None):143    global MATPLOTLIB_FLAG144    if not MATPLOTLIB_FLAG:145        import matplotlib146 147        matplotlib.use("Agg")148        MATPLOTLIB_FLAG = True149        mpl_logger = logging.getLogger("matplotlib")150        mpl_logger.setLevel(logging.WARNING)151    import matplotlib.pylab as plt152    import numpy as np153 154    fig, ax = plt.subplots(figsize=(6, 4))155    im = ax.imshow(156        alignment.transpose(), aspect="auto", origin="lower", interpolation="none"157    )158    fig.colorbar(im, ax=ax)159    xlabel = "Decoder timestep"160    if info is not None:161        xlabel += "\n\n" + info162    plt.xlabel(xlabel)163    plt.ylabel("Encoder timestep")164    plt.tight_layout()165 166    fig.canvas.draw()167    data = np.fromstring(fig.canvas.tostring_rgb(), dtype=np.uint8, sep="")168    data = data.reshape(fig.canvas.get_width_height()[::-1] + (3,))169    plt.close()170    return data171 172 173def load_wav_to_torch(full_path):174    data, sampling_rate = librosa.load(full_path, sr=None)175    return torch.FloatTensor(data), sampling_rate176 177 178def load_filepaths_and_text(filename, split="|"):179    with open(filename, encoding="utf-8") as f:180        filepaths_and_text = [line.strip().split(split) for line in f]181    return filepaths_and_text182 183 184def get_hparams(init=True, stage=1):185    parser = argparse.ArgumentParser()186    parser.add_argument(187        "-c",188        "--config",189        type=str,190        default="./configs/s2.json",191        help="JSON file for configuration",192    )193    parser.add_argument(194        "-p", "--pretrain", type=str, required=False, default=None, help="pretrain dir"195    )196    parser.add_argument(197        "-rs",198        "--resume_step",199        type=int,200        required=False,201        default=None,202        help="resume step",203    )204    # parser.add_argument('-e', '--exp_dir', type=str, required=False,default=None,help='experiment directory')205    # parser.add_argument('-g', '--pretrained_s2G', type=str, required=False,default=None,help='pretrained sovits gererator weights')206    # parser.add_argument('-d', '--pretrained_s2D', type=str, required=False,default=None,help='pretrained sovits discriminator weights')207 208    args = parser.parse_args()209 210    config_path = args.config211    with open(config_path, "r") as f:212        data = f.read()213    config = json.loads(data)214 215    hparams = HParams(**config)216    hparams.pretrain = args.pretrain217    hparams.resume_step = args.resume_step218    # hparams.data.exp_dir = args.exp_dir219    if stage == 1:220        model_dir = hparams.s1_ckpt_dir221    else:222        model_dir = hparams.s2_ckpt_dir223    config_save_path = os.path.join(model_dir, "config.json")224 225    if not os.path.exists(model_dir):226        os.makedirs(model_dir)227 228    with open(config_save_path, "w") as f:229        f.write(data)230    return hparams231 232 233def clean_checkpoints(path_to_models="logs/44k/", n_ckpts_to_keep=2, sort_by_time=True):234    """Freeing up space by deleting saved ckpts235 236    Arguments:237    path_to_models    --  Path to the model directory238    n_ckpts_to_keep   --  Number of ckpts to keep, excluding G_0.pth and D_0.pth239    sort_by_time      --  True -> chronologically delete ckpts240                          False -> lexicographically delete ckpts241    """242    import re243 244    ckpts_files = [245        f246        for f in os.listdir(path_to_models)247        if os.path.isfile(os.path.join(path_to_models, f))248    ]249    name_key = lambda _f: int(re.compile("._(\d+)\.pth").match(_f).group(1))250    time_key = lambda _f: os.path.getmtime(os.path.join(path_to_models, _f))251    sort_key = time_key if sort_by_time else name_key252    x_sorted = lambda _x: sorted(253        [f for f in ckpts_files if f.startswith(_x) and not f.endswith("_0.pth")],254        key=sort_key,255    )256    to_del = [257        os.path.join(path_to_models, fn)258        for fn in (x_sorted("G")[:-n_ckpts_to_keep] + x_sorted("D")[:-n_ckpts_to_keep])259    ]260    del_info = lambda fn: logger.info(f".. Free up space by deleting ckpt {fn}")261    del_routine = lambda x: [os.remove(x), del_info(x)]262    rs = [del_routine(fn) for fn in to_del]263 264 265def get_hparams_from_dir(model_dir):266    config_save_path = os.path.join(model_dir, "config.json")267    with open(config_save_path, "r") as f:268        data = f.read()269    config = json.loads(data)270 271    hparams = HParams(**config)272    hparams.model_dir = model_dir273    return hparams274 275 276def get_hparams_from_file(config_path):277    with open(config_path, "r") as f:278        data = f.read()279    config = json.loads(data)280 281    hparams = HParams(**config)282    return hparams283 284 285def check_git_hash(model_dir):286    source_dir = os.path.dirname(os.path.realpath(__file__))287    if not os.path.exists(os.path.join(source_dir, ".git")):288        logger.warn(289            "{} is not a git repository, therefore hash value comparison will be ignored.".format(290                source_dir291            )292        )293        return294 295    cur_hash = subprocess.getoutput("git rev-parse HEAD")296 297    path = os.path.join(model_dir, "githash")298    if os.path.exists(path):299        saved_hash = open(path).read()300        if saved_hash != cur_hash:301            logger.warn(302                "git hash values are different. {}(saved) != {}(current)".format(303                    saved_hash[:8], cur_hash[:8]304                )305            )306    else:307        open(path, "w").write(cur_hash)308 309 310def get_logger(model_dir, filename="train.log"):311    global logger312    logger = logging.getLogger(os.path.basename(model_dir))313    logger.setLevel(logging.DEBUG)314 315    formatter = logging.Formatter("%(asctime)s\t%(name)s\t%(levelname)s\t%(message)s")316    if not os.path.exists(model_dir):317        os.makedirs(model_dir)318    h = logging.FileHandler(os.path.join(model_dir, filename))319    h.setLevel(logging.DEBUG)320    h.setFormatter(formatter)321    logger.addHandler(h)322    return logger323 324 325class HParams:326    def __init__(self, **kwargs):327        for k, v in kwargs.items():328            if type(v) == dict:329                v = HParams(**v)330            self[k] = v331 332    def keys(self):333        return self.__dict__.keys()334 335    def items(self):336        return self.__dict__.items()337 338    def values(self):339        return self.__dict__.values()340 341    def __len__(self):342        return len(self.__dict__)343 344    def __getitem__(self, key):345        return getattr(self, key)346 347    def __setitem__(self, key, value):348        return setattr(self, key, value)349 350    def __contains__(self, key):351        return key in self.__dict__352 353    def __repr__(self):354        return self.__dict__.__repr__()355 356 357if __name__ == "__main__":358    print(359        load_wav_to_torch(360            "/home/fish/wenetspeech/dataset_vq/Y0000022499_wHFSeHEx9CM/S00261.flac"361        )362    )363