train/encode_audio.py

import time
import json
import torch
import multiprocessing
from tqdm import tqdm
import concurrent.futures
from model import PromptCondAudioDiffusion
from diffusers import DDIMScheduler, DDPMScheduler
import torchaudio
import librosa
import os
import math
import numpy as np
from tools.get_melvaehifigan48k import build_pretrained_models
import tools.torch_tools as torch_tools
from safetensors.torch import load_file
import subprocess

def get_free_gpu() -> int:
    """Return the GPU ID with the least memory usage"""
    cmd = "nvidia-smi --query-gpu=index,memory.free --format=csv,noheader,nounits"
    result = subprocess.check_output(cmd.split()).decode().strip().split("\n")
    
    free_list = []
    for line in result:
        idx, free_mem = line.split(",")
        free_list.append((int(idx), int(free_mem)))  # (GPU id, free memory MiB)
    
    # Sort by remaining memory
    free_list.sort(key=lambda x: x[1], reverse=True)
    return free_list[0][0]

SAVE_DIR = "xxx/suno_mucodec_en"
DATA_PATH = "xxx/paths_en.jsonl"
DEVICE = f"cuda:{get_free_gpu()}"
print(f"Use {DEVICE}")

class MuCodec:
    def __init__(self, \
        model_path, \
        layer_num, \
        load_main_model=True):
        
        self.layer_num = layer_num - 1
        self.sample_rate = 48000
        self.device = DEVICE

        self.MAX_DURATION = 360
        if load_main_model:
            audio_ldm_path = os.path.dirname(os.path.abspath(__file__)) + "/tools/audioldm_48k.pth"
            self.vae, self.stft = build_pretrained_models(audio_ldm_path)
            self.vae, self.stft = self.vae.eval().to(DEVICE), self.stft.eval().to(DEVICE)
            main_config = {
                "num_channels":32,
                "unet_model_name":None,
                "unet_model_config_path":os.path.dirname(os.path.abspath(__file__)) + "/configs/models/transformer2D.json",
                "snr_gamma":None,
            }
            self.model = PromptCondAudioDiffusion(**main_config)
            if model_path.endswith('.safetensors'):
                main_weights = load_file(model_path)
            else:
                main_weights = torch.load(model_path, map_location='cpu')
            self.model.load_state_dict(main_weights, strict=False)
            self.model = self.model.to(DEVICE)
            print ("Successfully loaded checkpoint from:", model_path)
        else:
            main_config = {
                "num_channels":32,
                "unet_model_name":None,
                "unet_model_config_path":None,
                "snr_gamma":None,
            }
            self.model = PromptCondAudioDiffusion(**main_config).to(DEVICE)
            main_weights = torch.load(model_path, map_location='cpu')
            self.model.load_state_dict(main_weights, strict=False)
            self.model = self.model.to(DEVICE)
            print ("Successfully loaded checkpoint from:", model_path)
        
        self.model.eval()
        self.model.init_device_dtype(torch.device(DEVICE), torch.float32)
        print("scaling factor: ", self.model.normfeat.std)

    def file2code(self, fname):
        orig_samples, fs = torchaudio.load(fname)
        if(fs!=self.sample_rate):
            orig_samples = torchaudio.functional.resample(orig_samples, fs, self.sample_rate)
            fs = self.sample_rate
        if orig_samples.shape[0] == 1:
            orig_samples = torch.cat([orig_samples, orig_samples], 0)
        return self.sound2code(orig_samples)

    @torch.no_grad()
    @torch.autocast(device_type="cuda", dtype=torch.float32)
    def sound2code(self, orig_samples, batch_size=3):
        if(orig_samples.ndim == 2):
            audios = orig_samples.unsqueeze(0).to(self.device)
        elif(orig_samples.ndim == 3):
            audios = orig_samples.to(self.device)
        else:
            assert orig_samples.ndim in (2,3), orig_samples.shape
        audios = self.preprocess_audio(audios)
        audios = audios.squeeze(0)
        orig_length = audios.shape[-1]
        min_samples = int(40.96 * self.sample_rate)
        output_len = int(orig_length / float(self.sample_rate) * 25) + 1
        print("output_len: ", output_len)

        while(audios.shape[-1] < min_samples + 480):
            audios = torch.cat([audios, audios], -1)
        int_max_len=audios.shape[-1]//min_samples+1
        # print("int_max_len: ", int_max_len)
        audios = torch.cat([audios, audios], -1)
        # print("audios:",audios.shape)
        audios=audios[:,:int(int_max_len*(min_samples+480))]
        codes_list=[]

        audio_input = audios.reshape(2, -1, min_samples+480).permute(1, 0, 2).reshape(-1, 2, min_samples+480)

        for audio_inx in range(0, audio_input.shape[0], batch_size):
            # import pdb; pdb.set_trace()
            codes, _, spk_embeds = self.model.fetch_codes_batch((audio_input[audio_inx:audio_inx+batch_size]), additional_feats=[],layer=self.layer_num)
            codes_list.append(torch.cat(codes, 1))
            # print("codes_list",codes_list[0].shape)

        codes = torch.cat(codes_list, 0).permute(1,0,2).reshape(1, -1)[None] # B 3 T -> 3 B T
        codes=codes[:,:,:output_len]

        return codes

    @torch.no_grad()
    def code2sound(self, codes, prompt=None, duration=40.96, guidance_scale=1.5, num_steps=20, disable_progress=False):
        codes = codes.to(self.device)
        first_latent = torch.randn(codes.shape[0], 32, 512, 32).to(self.device)
        first_latent_length = 0
        first_latent_codes_length = 0
        if(isinstance(prompt, torch.Tensor)):
            prompt = prompt.to(self.device)
            if(prompt.ndim == 3):
                assert prompt.shape[0] == 1, prompt.shape
                prompt = prompt[0]
            elif(prompt.ndim == 1):
                prompt = prompt.unsqueeze(0).repeat(2,1)
            elif(prompt.ndim == 2):
                if(prompt.shape[0] == 1):
                    prompt = prompt.repeat(2,1)

            if(prompt.shape[-1] < int(30.76 * self.sample_rate)):
                prompt = prompt[:,:int(10.24*self.sample_rate)] # limit max length to 10.24
            else:
                prompt = prompt[:,int(20.48*self.sample_rate):int(30.72*self.sample_rate)] # limit max length to 10.24
            
            true_mel , _, _ = torch_tools.wav_to_fbank2(prompt, -1, fn_STFT=self.stft) # maximum 10.24s
            true_mel = true_mel.unsqueeze(1).to(self.device)
            true_latent = torch.cat([self.vae.get_first_stage_encoding(self.vae.encode_first_stage(true_mel[[m]])) for m in range(true_mel.shape[0])],0)
            true_latent = true_latent.reshape(true_latent.shape[0]//2, -1, true_latent.shape[2], true_latent.shape[3]).detach()
            
            first_latent[:,:,0:true_latent.shape[2],:] = true_latent
            first_latent_length = true_latent.shape[2]
            first_latent_codes = self.sound2code(prompt)[:,:,0:first_latent_length*2] # B 4 T
            first_latent_codes_length = first_latent_codes.shape[-1]
            codes = torch.cat([first_latent_codes, codes], -1)

        min_samples = 1024
        hop_samples = min_samples // 4 * 3
        ovlp_samples = min_samples - hop_samples
        hop_frames = hop_samples // 2
        ovlp_frames = ovlp_samples // 2

        codes_len= codes.shape[-1]
        target_len = int((codes_len - first_latent_codes_length) / 100 * 4 * self.sample_rate)

        if(codes_len < min_samples):
            while(codes.shape[-1] < min_samples):
                codes = torch.cat([codes, codes], -1)
            codes = codes[:,:,0:min_samples]
        codes_len = codes.shape[-1]
        if((codes_len - ovlp_frames) % hop_samples > 0):
            len_codes=math.ceil((codes_len - ovlp_samples) / float(hop_samples)) * hop_samples + ovlp_samples
            while(codes.shape[-1] < len_codes):
                codes = torch.cat([codes, codes], -1)
            codes = codes[:,:,0:len_codes]
        latent_length = 512
        latent_list = []
        spk_embeds = torch.zeros([1, 32, 1, 32], device=codes.device)
        with torch.autocast(device_type="cuda", dtype=torch.float16):
            for sinx in range(0, codes.shape[-1]-hop_samples, hop_samples):
                codes_input=[]
                codes_input.append(codes[:,:,sinx:sinx+min_samples])
                if(sinx == 0):
                    incontext_length = first_latent_length
                    latents = self.model.inference_codes(codes_input, spk_embeds, first_latent, latent_length, incontext_length, additional_feats=[], guidance_scale=1.5, num_steps = num_steps, disable_progress=disable_progress, scenario='other_seg')
                    latent_list.append(latents)
                else:
                    true_latent = latent_list[-1][:,:,-ovlp_frames:,:]
                    len_add_to_512 = 512 - true_latent.shape[-2]
                    incontext_length = true_latent.shape[-2]
                    true_latent = torch.cat([true_latent, torch.randn(true_latent.shape[0], true_latent.shape[1], len_add_to_512, true_latent.shape[-1]).to(self.device)], -2)
                    latents = self.model.inference_codes(codes_input, spk_embeds, true_latent, latent_length, incontext_length,  additional_feats=[], guidance_scale=1.5, num_steps = num_steps, disable_progress=disable_progress, scenario='other_seg')
                    latent_list.append(latents)

        latent_list = [l.float() for l in latent_list]
        latent_list[0] = latent_list[0][:,:,first_latent_length:,:]
        min_samples =  int(duration * self.sample_rate)
        hop_samples = min_samples // 4 * 3
        ovlp_samples = min_samples - hop_samples
        with torch.no_grad():
            output = None
            for i in range(len(latent_list)):
                latent = latent_list[i]
                bsz , ch, t, f = latent.shape
                latent = latent.reshape(bsz*2, ch//2, t, f)
                mel = self.vae.decode_first_stage(latent)
                cur_output = self.vae.decode_to_waveform(mel)
                cur_output = torch.from_numpy(cur_output)[:, 0:min_samples]

                if output is None:
                    output = cur_output
                else:
                    ov_win = torch.from_numpy(np.linspace(0, 1, ovlp_samples)[None, :])
                    ov_win = torch.cat([ov_win, 1 - ov_win], -1)
                    output[:, -ovlp_samples:] = output[:, -ovlp_samples:] * ov_win[:, -ovlp_samples:] + cur_output[:, 0:ovlp_samples] * ov_win[:, 0:ovlp_samples]
                    output = torch.cat([output, cur_output[:, ovlp_samples:]], -1)
            output = output[:, 0:target_len]
        return output

    @torch.no_grad()
    def preprocess_audio(self, input_audios, threshold=0.8):
        assert len(input_audios.shape) == 3, input_audios.shape
        nchan = input_audios.shape[1]
        input_audios = input_audios.reshape(input_audios.shape[0], -1)
        norm_value = torch.ones_like(input_audios[:,0])
        max_volume = input_audios.abs().max(dim=-1)[0]
        norm_value[max_volume>threshold] = max_volume[max_volume>threshold] / threshold
        return input_audios.reshape(input_audios.shape[0], nchan, -1)/norm_value.unsqueeze(-1).unsqueeze(-1)
    
    @torch.no_grad()
    def sound2sound(self, sound, prompt=None, min_duration=40.96, steps=50, disable_progress=False):
        start_time = time.time()
        codes = self.sound2code(sound)
        mid_time = time.time()
        elapsed_1 = mid_time - start_time
        print(f"sound2code: {elapsed_1:.3f}s")
        wave = self.code2sound(codes, prompt, duration=min_duration, guidance_scale=1.5, num_steps=steps, disable_progress=disable_progress)
        end_time = time.time()
        elapsed_2 = end_time - mid_time
        print(f"code2sound: step-{steps}, {elapsed_2:.3f}s")
        return wave

def load_jsonl(path:str) -> list[dict]:
    dataset = []
    with open(path, 'r') as file:
        for line in tqdm(file, desc=f"Loading {path}"):
            dataset.append(json.loads(line))
    return dataset

ckpt_path = os.path.join(os.path.dirname(os.path.abspath(__file__)), "ckpt/mucodec.pt")
mucodec = MuCodec(model_path=ckpt_path,layer_num=7,load_main_model=True)

def pure_name(path:str):
    """Get the original name of a file path (without extension)"""
    basename = os.path.basename(path)
    dot_pos = basename.rfind('.')
    if dot_pos == -1:
        return basename
    return basename[:dot_pos]

def music_encode(path:str):
    # Process to get path after vocal separation
    name = pure_name(path)
    # path = path.replace("enhanced", "separated").replace(".mp3", "/vocals.wav")

    sound, fs = torchaudio.load(path)
    if(fs!=48000):
        sound = torchaudio.functional.resample(sound, fs, 48000)
    if(sound.shape[0]==1):
        sound = torch.cat([sound, sound],0)
    code = mucodec.sound2code(sound)

    # save_path = f"xxx/seperate_codes/{name}.pt"
    save_path = os.path.join(SAVE_DIR, f"{name}.pt")
    torch.save(code, save_path)
    return save_path

# ls xxx/seperate_codes/ -l | grep -c '^-'
def remove_finish(dataset:list[dict]) -> list[dict]:
    """Remove already encoded items from dataset"""
    names = []
    for path in os.listdir(SAVE_DIR):
        names.append(pure_name(path))
    new_dataset = []
    dup_num = 0
    for line in tqdm(dataset):
        name = pure_name(line['path'])
        if name not in names:
            new_dataset.append(line)
        else:
            dup_num += 1
    print(f"Remove dup num: {dup_num}")
    return new_dataset

if __name__=="__main__":
    multiprocessing.set_start_method('spawn', force=True)
    dataset = load_jsonl(DATA_PATH)
    dataset = remove_finish(dataset)

    with concurrent.futures.ProcessPoolExecutor(max_workers=4) as executor:
        # Submit to task pool
        futures = [executor.submit(music_encode, line['path']) for line in dataset]
        # Progress bar display
        for future in tqdm(concurrent.futures.as_completed(futures),
                total=len(dataset),
                desc="Encoding audio count",
                unit="files",
                ncols=100,
                bar_format="{l_bar}{bar}| {n_fmt}/{total_fmt} [{elapsed}<{remaining}]"):
            try:
                result = future.result()  # Explicitly get result (must call even if not needed)
            except Exception as e:
                print(f"Task failed: {e}")  # Catch exception to avoid interrupting progress bar