Merge pull request #331 from svc-develop-team/add-multi-GPU-preproces…

…s-support

Add multi gpu preprocess support

README.md

@@ -258,6 +258,15 @@ Add `--vol_aug` if you want to enable loudness embedding:
 python preprocess_flist_config.py --speech_encoder vec768l12 --vol_aug
 ```
 
+**Speed Up preprocess**
+
+If your dataset is pretty large,you can increase the param `--num_processes` like that:
+
+```shell
+python preprocess_flist_config.py --speech_encoder vec768l12 --vol_aug --num_processes 8
+```
+All the worker will be assigned to different GPU if you have more than one GPUs.
+
 After enabling loudness embedding, the trained model will match the loudness of the input source; otherwise, it will match the loudness of the training set.
 
 #### You can modify some parameters in the generated config.json and diffusion.yaml

README_zh_CN.md

@@ -260,6 +260,12 @@ wavlmbase+
 python preprocess_flist_config.py --speech_encoder vec768l12 --vol_aug
 ```
 
+**加速预处理**
+如若您的数据集比较大，可以尝试添加`--num_processes`参数：
+```shell
+python preprocess_flist_config.py --speech_encoder vec768l12 --vol_aug --num_processes 8
+```
+所有的Workers会被自动分配到多个GPU上（如果您有多个GPU的话）
 使用后训练出的模型将匹配到输入源响度，否则为训练集响度。
 
 #### 此时可以在生成的 config.json 与 diffusion.yaml 修改部分参数

preprocess_hubert_f0.py

@@ -1,6 +1,6 @@
 import argparse
 import logging
-import multiprocessing
+import torch.multiprocessing as mp
 import os
 import random
 from concurrent.futures import ProcessPoolExecutor
@@ -106,10 +106,14 @@ def process_one(filename, hmodel,f0p,diff=False,mel_extractor=None):
 
 def process_batch(file_chunk, f0p, diff=False, mel_extractor=None):
     print("Loading speech encoder for content...")
-    device = "cuda" if torch.cuda.is_available() else "cpu"
+    rank = mp.current_process()._identity
+    rank = rank[0] if len(rank) > 0 else 0
+    if torch.cuda.is_available():
+        gpu_id = rank % torch.cuda.device_count()
+        device = torch.device(f"cuda:{gpu_id}")
+    print("Rank {rank} uses device {device}")
     hmodel = utils.get_speech_encoder(speech_encoder, device=device)
     print("Loaded speech encoder.")
-
     for filename in tqdm(file_chunk):
         process_one(filename, hmodel, f0p, diff, mel_extractor)
 
@@ -121,7 +125,6 @@ def parallel_process(filenames, num_processes, f0p, diff, mel_extractor):
             end = int((i + 1) * len(filenames) / num_processes)
             file_chunk = filenames[start:end]
             tasks.append(executor.submit(process_batch, file_chunk, f0p, diff, mel_extractor))
-
         for task in tqdm(tasks):
             task.result()
 
@@ -139,7 +142,6 @@ def parallel_process(filenames, num_processes, f0p, diff, mel_extractor):
     parser.add_argument(
         '--num_processes', type=int, default=1, help='You are advised to set the number of processes to the same as the number of CPU cores'
     )
-    device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
     args = parser.parse_args()
     f0p = args.f0_predictor
     print(speech_encoder)
@@ -148,16 +150,16 @@ def parallel_process(filenames, num_processes, f0p, diff, mel_extractor):
     if args.use_diff:
         print("use_diff")
         print("Loading Mel Extractor...")
-        mel_extractor = Vocoder(dconfig.vocoder.type, dconfig.vocoder.ckpt, device = device)
+        mel_extractor = Vocoder(dconfig.vocoder.type, dconfig.vocoder.ckpt, device = "cuda:0")
         print("Loaded Mel Extractor.")
     else:
         mel_extractor = None
     filenames = glob(f"{args.in_dir}/*/*.wav", recursive=True)  # [:10]
     shuffle(filenames)
-    multiprocessing.set_start_method("spawn", force=True)
+    mp.set_start_method("spawn", force=True)
 
     num_processes = args.num_processes
     if num_processes == 0:
         num_processes = os.cpu_count()
 
-    parallel_process(filenames, num_processes, f0p, args.use_diff, mel_extractor)
+    parallel_process(filenames, num_processes, f0p, args.use_diff, mel_extractor)

utils.py

@@ -43,7 +43,6 @@ def normalize_f0(f0, x_mask, uv, random_scale=True):
     if torch.isnan(f0_norm).any():
         exit(0)
     return f0_norm * x_mask
-
 def plot_data_to_numpy(x, y):
     global MATPLOTLIB_FLAG
     if not MATPLOTLIB_FLAG: