dawood's picture
dawood HF staff
Duplicate from haoheliu/audioldm-text-to-audio-generation
905f171
raw
history blame
3.83 kB
import importlib
from inspect import isfunction
import os
import soundfile as sf
def seed_everything(seed):
import random, os
import numpy as np
import torch
random.seed(seed)
os.environ['PYTHONHASHSEED'] = str(seed)
np.random.seed(seed)
torch.manual_seed(seed)
torch.cuda.manual_seed(seed)
torch.backends.cudnn.deterministic = True
torch.backends.cudnn.benchmark = True
def save_wave(waveform, savepath, name="outwav"):
if type(name) is not list:
name = [name] * waveform.shape[0]
for i in range(waveform.shape[0]):
path = os.path.join(
savepath,
"%s_%s.wav"
% (
os.path.basename(name[i])
if (not ".wav" in name[i])
else os.path.basename(name[i]).split(".")[0],
i,
),
)
sf.write(path, waveform[i, 0], samplerate=16000)
def exists(x):
return x is not None
def default(val, d):
if exists(val):
return val
return d() if isfunction(d) else d
def count_params(model, verbose=False):
total_params = sum(p.numel() for p in model.parameters())
if verbose:
print(f"{model.__class__.__name__} has {total_params * 1.e-6:.2f} M params.")
return total_params
def get_obj_from_str(string, reload=False):
module, cls = string.rsplit(".", 1)
if reload:
module_imp = importlib.import_module(module)
importlib.reload(module_imp)
return getattr(importlib.import_module(module, package=None), cls)
def instantiate_from_config(config):
if not "target" in config:
if config == "__is_first_stage__":
return None
elif config == "__is_unconditional__":
return None
raise KeyError("Expected key `target` to instantiate.")
return get_obj_from_str(config["target"])(**config.get("params", dict()))
def default_audioldm_config():
return {'wave_file_save_path': './output', 'id': {'version': 'v1', 'name': 'default', 'root': '/mnt/fast/nobackup/users/hl01486/projects/general_audio_generation/AudioLDM-python/config/default/latent_diffusion.yaml'}, 'model': {'device': 'cuda', 'reload_from_ckpt': '/mnt/fast/nobackup/scratch4weeks/hl01486/exps/audio_generation/stablediffusion/LDM/audioverse/2023_01_14_full_F4_B_spatial_v2_v1/checkpoints/last.ckpt', 'target': 'audioldm.pipline.LatentDiffusion', 'params': {'base_learning_rate': 5e-06, 'linear_start': 0.0015, 'linear_end': 0.0195, 'num_timesteps_cond': 1, 'log_every_t': 200, 'timesteps': 1000, 'first_stage_key': 'fbank', 'cond_stage_key': 'waveform', 'latent_t_size': 256, 'latent_f_size': 16, 'channels': 8, 'cond_stage_trainable': True, 'conditioning_key': 'film', 'monitor': 'val/loss_simple_ema', 'scale_by_std': True, 'unet_config': {'target': 'audioldm.latent_diffusion.openaimodel.UNetModel', 'params': {'image_size': 64, 'extra_film_condition_dim': 512, 'extra_film_use_concat': True, 'in_channels': 8, 'out_channels': 8, 'model_channels': 128, 'attention_resolutions': [8, 4, 2], 'num_res_blocks': 2, 'channel_mult': [1, 2, 3, 5], 'num_head_channels': 32, 'use_spatial_transformer': True}}, 'first_stage_config': {'base_learning_rate': 4.5e-05, 'target': 'audioldm.variational_autoencoder.autoencoder.AutoencoderKL', 'params': {'monitor': 'val/rec_loss', 'image_key': 'fbank', 'subband': 1, 'embed_dim': 8, 'time_shuffle': 1, 'ddconfig': {'double_z': True, 'z_channels': 8, 'resolution': 256, 'downsample_time': False, 'in_channels': 1, 'out_ch': 1, 'ch': 128, 'ch_mult': [1, 2, 4], 'num_res_blocks': 2, 'attn_resolutions': [], 'dropout': 0.0}}}, 'cond_stage_config': {'target': 'audioldm.clap.encoders.CLAPAudioEmbeddingClassifierFreev2', 'params': {'key': 'waveform', 'sampling_rate': 16000, 'embed_mode': 'audio', 'unconditional_prob': 0.1}}}}}