File size: 10,078 Bytes
2b5b9ef
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
import argparse, os, sys, glob
import pathlib
directory = pathlib.Path(os.getcwd())
print(directory)
sys.path.append(str(directory))
import torch
import numpy as np
from omegaconf import OmegaConf
from PIL import Image
from tqdm import tqdm, trange
from ldm.util import instantiate_from_config
from ldm.models.diffusion.ddim import DDIMSampler
from ldm.models.diffusion.plms import PLMSSampler
import pandas as pd
from torch.utils.data import DataLoader 
from tqdm import tqdm
from icecream import ic
from pathlib import Path
import yaml
from vocoder.bigvgan.models import VocoderBigVGAN
import soundfile
# from pytorch_memlab import LineProfiler,profile

def load_model_from_config(config, ckpt = None, verbose=True):
    model = instantiate_from_config(config.model)
    if ckpt:
        print(f"Loading model from {ckpt}")
        pl_sd = torch.load(ckpt, map_location="cpu")
        sd = pl_sd["state_dict"]
        
        m, u = model.load_state_dict(sd, strict=False)
        if len(m) > 0 and verbose:
            print("missing keys:")
            print(m)
        if len(u) > 0 and verbose:
            print("unexpected keys:")
            print(u)
    else:
        print(f"Note chat no ckpt is loaded !!!")

    model.cuda()
    model.eval()
    return model


def parse_args():
    parser = argparse.ArgumentParser()

    parser.add_argument(
        "--prompt_txt",
        type=str,
        nargs="?",
        default="prompt.txt",
        help="txt file with prompts in it"
    )
    parser.add_argument(
        "--sample_rate",
        type=int,
        default="22050",
        help="sample rate of wav"
    )
    parser.add_argument(
        "--inpaint",
        action='store_true',
        help="if test txt guided inpaint task"
    )
    parser.add_argument(
        "--test-dataset",
        default="none",
        help="test which dataset: audiocaps/clotho/fsd50k"
    )
    parser.add_argument(
        "--outdir",
        type=str,
        nargs="?",
        help="dir to write results to",
        default="outputs/txt2audio-samples"
    )
    parser.add_argument(
        "--ddim_steps",
        type=int,
        default=100,
        help="number of ddim sampling steps",
    )

    parser.add_argument(
        "--plms",
        action='store_true',
        help="use plms sampling",
    )

    parser.add_argument(
        "--ddim_eta",
        type=float,
        default=0.0,
        help="ddim eta (eta=0.0 corresponds to deterministic sampling",
    )
    parser.add_argument(
        "--n_iter",
        type=int,
        default=1,
        help="sample this often",
    )

    parser.add_argument(
        "--H",
        type=int,
        default=80,
        help="image height, in pixel space",
    )

    parser.add_argument(
        "--W",
        type=int,
        default=848,
        help="image width, in pixel space",
    )

    parser.add_argument(
        "--n_samples",
        type=int,
        default=1,
        help="how many samples to produce for the given prompt",
    )

    parser.add_argument(
        "--scale",
        type=float,
        default=5.0, # if it's 1, only condition is taken into consideration
        help="unconditional guidance scale: eps = eps(x, empty) + scale * (eps(x, cond) - eps(x, empty))",
    )
    parser.add_argument(
        "-r",
        "--resume",
        type=str,
        const=True,
        default="",
        nargs="?",
        help="resume from logdir or checkpoint in logdir",
    )
    parser.add_argument(
        "-b",
        "--base",
        type=str,
        help="paths to base configs. Loaded from left-to-right. "
             "Parameters can be overwritten or added with command-line options of the form `--key value`.",
        default="",
    )
    parser.add_argument(
        "--vocoder-ckpt",
        type=str,
        help="paths to vocoder checkpoint",
        default='vocoder/logs/audioset',
    )

    return parser.parse_args()

class GenSamples:
    def __init__(self,opt,sampler,model,outpath,vocoder = None,save_mel = True,save_wav = True) -> None:
        self.opt = opt
        self.sampler = sampler
        self.model = model
        self.outpath = outpath
        if save_wav:
            assert vocoder is not None
            self.vocoder = vocoder
        self.save_mel = save_mel
        self.save_wav = save_wav
        self.channel_dim = self.model.channels
    
    def gen_test_sample(self,prompt,mel_name = None,wav_name = None):# prompt is {'ori_caption':’xxx‘,'struct_caption':'xxx'}
        uc = None
        record_dicts = []
        # if os.path.exists(os.path.join(self.outpath,mel_name+f'_0.npy')):
        #     return record_dicts
        if self.opt.scale != 1.0:
            emptycap = {'ori_caption':self.opt.n_samples*[""],'struct_caption':self.opt.n_samples*[""]}
            uc = self.model.get_learned_conditioning(emptycap)
        for n in range(self.opt.n_iter):# trange(self.opt.n_iter, desc="Sampling"):
            for k,v in prompt.items():
                prompt[k] = self.opt.n_samples * [v]
            c = self.model.get_learned_conditioning(prompt)# shape:[1,77,1280],即还没有变成句子embedding,仍是每个单词的embedding
            if self.channel_dim>0:
                shape = [self.channel_dim, self.opt.H, self.opt.W]  # (z_dim, 80//2^x, 848//2^x)
            else:
                shape = [self.opt.H, self.opt.W]
            samples_ddim, _ = self.sampler.sample(S=self.opt.ddim_steps,
                                                conditioning=c,
                                                batch_size=self.opt.n_samples,
                                                shape=shape,
                                                verbose=False,
                                                unconditional_guidance_scale=self.opt.scale,
                                                unconditional_conditioning=uc,
                                                # quantize_x0=use_quantize,
                                                eta=self.opt.ddim_eta)

            x_samples_ddim = self.model.decode_first_stage(samples_ddim)

            for idx,spec in enumerate(x_samples_ddim):
                spec = spec.squeeze(0).cpu().numpy()
                record_dict = {'caption':prompt['ori_caption'][0]}
                if self.save_mel:
                    mel_path = os.path.join(self.outpath,mel_name+f'_{idx}.npy')
                    np.save(mel_path,spec)
                    record_dict['mel_path'] = mel_path
                if self.save_wav:
                    wav = self.vocoder.vocode(spec)
                    wav_path = os.path.join(self.outpath,wav_name+f'_{idx}.wav')
                    soundfile.write(wav_path, wav, self.opt.sample_rate)
                    record_dict['audio_path'] = wav_path
                record_dicts.append(record_dict)
        return record_dicts


def main():
    opt = parse_args()

    config = OmegaConf.load(opt.base)
    # print("-------quick debug no load ckpt---------")
    # model = instantiate_from_config(config['model'])# for quick debug
    model = load_model_from_config(config, opt.resume)

    device = torch.device("cuda") if torch.cuda.is_available() else torch.device("cpu")
    model = model.to(device)

    if opt.plms:
        sampler = PLMSSampler(model)
    else:
        sampler = DDIMSampler(model)

    os.makedirs(opt.outdir, exist_ok=True)
    if 'mel' in opt.vocoder_ckpt:
        vocoder = VocoderMelGan(opt.vocoder_ckpt,device)
    elif 'hifi' in opt.vocoder_ckpt:
        vocoder = VocoderHifigan(opt.vocoder_ckpt,device)
    elif 'bigv' in opt.vocoder_ckpt:
        vocoder = VocoderBigVGAN(opt.vocoder_ckpt,device)


    generator = GenSamples(opt,sampler,model,opt.outdir,vocoder,save_mel = False,save_wav = True)
    csv_dicts = []
    
    with torch.no_grad():
        with model.ema_scope():
            if opt.test_dataset != 'none':
                if opt.test_dataset == 'audiocaps':
                    test_dataset = instantiate_from_config(config['test_dataset'])
                elif opt.test_dataset == 'clotho':
                    test_dataset = instantiate_from_config(config['test_dataset2'])
                elif opt.test_dataset == 'fsd50k':
                    test_dataset = instantiate_from_config(config['test_dataset3'])
                elif opt.test_dataset == 'musiccap':
                    test_dataset = instantiate_from_config(config['test_dataset'])
                print(f"Dataset: {type(test_dataset)} LEN: {len(test_dataset)}")
                for item in tqdm(test_dataset):
                    import ipdb
                    # ipdb.set_trace()
                    prompt,f_name = item['caption'],item['f_name']
                    vname_num_split_index = f_name.rfind('_')# file_names[b]:video_name+'_'+num
                    v_n,num = f_name[:vname_num_split_index],f_name[vname_num_split_index+1:]
                    mel_name = f'{v_n}_sample_{num}'
                    wav_name = f'{v_n}_sample_{num}'
                    # write_gt_wav(v_n,opt.test_dataset2,opt.outdir,opt.sample_rate)
                    csv_dicts.extend(generator.gen_test_sample(prompt,mel_name=mel_name,wav_name=wav_name))
    
                df = pd.DataFrame.from_dict(csv_dicts)
                df.to_csv(os.path.join(opt.outdir,'result.csv'),sep='\t',index=False)
            else:
                with open(opt.prompt_txt,'r') as f:
                    prompts = f.readlines()
                for prompt in prompts:
                    wav_name = f'{prompt.strip().replace(" ", "-")}'
                    generator.gen_test_sample(prompt,wav_name=wav_name)

    print(f"Your samples are ready and waiting four you here: \n{opt.outdir} \nEnjoy.")

if __name__ == "__main__":
    main()