feat: update to latest torch & gradio version

Author: fumiama

.github/workflows/unitest.yml

@@ -21,7 +21,6 @@ jobs:
         wget https://github.com/fumiama/RVC-Models-Downloader/releases/download/v0.2.5/rvcmd_linux_amd64.deb
         sudo apt -y install ./rvcmd_linux_amd64.deb
         pip install --force pip==24.0 # fix fairseq installing issue https://github.com/facebookresearch/fairseq/issues/5552
-        python -m pip install --upgrade pip
         python -m pip install --upgrade setuptools
         python -m pip install --upgrade wheel
         pip install torch torchvision torchaudio

gui.py

@@ -251,7 +251,7 @@ def launcher(self):
                                 sg.FileBrowse(
                                     i18n("Select the .pth file"),
                                     initial_folder=os.path.join(
-                                        os.getcwd(), "assets/weights"
+                                        os.getcwd(), "assets", "weights"
                                     ),
                                     file_types=[("Model File", "*.pth")],
                                 ),

infer/lib/train/data_utils.py

@@ -111,7 +111,7 @@ def get_audio(self, filename):
         spec_filename = filename.replace(".wav", ".spec.pt")
         if os.path.exists(spec_filename):
             try:
-                spec = torch.load(spec_filename)
+                spec = torch.load(spec_filename, weights_only=True)
             except:
                 logger.warning("%s %s", spec_filename, traceback.format_exc())
                 spec = spectrogram_torch(

infer/lib/train/utils.py

@@ -71,9 +71,7 @@ def go(model, bkey):
 
 def load_checkpoint(checkpoint_path, model, optimizer=None, load_opt=1):
     assert os.path.isfile(checkpoint_path)
-    checkpoint_dict = torch.load(checkpoint_path, map_location="cpu")
-
-    saved_state_dict = checkpoint_dict["model"]
+    saved_state_dict = torch.load(checkpoint_path, map_location="cpu", weights_only=True)["model"]
     if hasattr(model, "module"):
         state_dict = model.module.state_dict()
     else:

infer/modules/train/train.py

@@ -131,9 +131,14 @@ def run(rank, n_gpus, hps: utils.HParams, logger: logging.Logger):
         writer = SummaryWriter(log_dir=hps.model_dir)
         writer_eval = SummaryWriter(log_dir=os.path.join(hps.model_dir, "eval"))
 
-    dist.init_process_group(
-        backend="gloo", init_method="env://", world_size=n_gpus, rank=rank
-    )
+    try:
+        dist.init_process_group(
+            backend="gloo" if os.name == "nt" or not torch.cuda.is_available() else "nccl", init_method="env://", world_size=n_gpus, rank=rank
+        )
+    except:
+        dist.init_process_group(
+            backend="gloo" if os.name == "nt" or not torch.cuda.is_available() else "nccl", init_method="env://?use_libuv=False", world_size=n_gpus, rank=rank
+        )
     torch.manual_seed(hps.train.seed)
     if torch.cuda.is_available():
         torch.cuda.set_device(rank)
@@ -238,13 +243,13 @@ def run(rank, n_gpus, hps: utils.HParams, logger: logging.Logger):
             if hasattr(net_g, "module"):
                 logger.info(
                     net_g.module.load_state_dict(
-                        torch.load(hps.pretrainG, map_location="cpu")["model"]
+                        torch.load(hps.pretrainG, map_location="cpu", weights_only=True)["model"]
                     )
                 )  ##测试不加载优化器
             else:
                 logger.info(
                     net_g.load_state_dict(
-                        torch.load(hps.pretrainG, map_location="cpu")["model"]
+                        torch.load(hps.pretrainG, map_location="cpu", weights_only=True)["model"]
                     )
                 )  ##测试不加载优化器
         if hps.pretrainD != "":
@@ -253,13 +258,13 @@ def run(rank, n_gpus, hps: utils.HParams, logger: logging.Logger):
             if hasattr(net_d, "module"):
                 logger.info(
                     net_d.module.load_state_dict(
-                        torch.load(hps.pretrainD, map_location="cpu")["model"]
+                        torch.load(hps.pretrainD, map_location="cpu", weights_only=True)["model"]
                     )
                 )
             else:
                 logger.info(
                     net_d.load_state_dict(
-                        torch.load(hps.pretrainD, map_location="cpu")["model"]
+                        torch.load(hps.pretrainD, map_location="cpu", weights_only=True)["model"]
                     )
                 )
 

infer/modules/vc/utils.py

@@ -1,4 +1,4 @@
-import os
+import os, pathlib
 
 from fairseq import checkpoint_utils
 
@@ -8,7 +8,7 @@ def get_index_path_from_model(sid):
         (
             f
             for f in [
-                os.path.join(root, name)
+                str(pathlib.Path(root, name))
                 for path in [os.getenv("outside_index_root"), os.getenv("index_root")]
                 for root, _, files in os.walk(path, topdown=False)
                 for name in files

rvc/f0/models.py

@@ -7,8 +7,9 @@ def get_rmvpe(
     from rvc.f0.e2e import E2E
 
     model = E2E(4, 1, (2, 2))
-    ckpt = torch.load(model_path, map_location=device)
+    ckpt = torch.load(model_path, map_location=device, weights_only=True)
     model.load_state_dict(ckpt)
+    del ckpt
     model.eval()
     if is_half:
         model = model.half()

web.py

@@ -46,7 +46,7 @@
 shutil.rmtree(tmp, ignore_errors=True)
 os.makedirs(tmp, exist_ok=True)
 os.makedirs(os.path.join(now_dir, "logs"), exist_ok=True)
-os.makedirs(os.path.join(now_dir, "assets/weights"), exist_ok=True)
+os.makedirs(os.path.join(now_dir, "assets", "weights"), exist_ok=True)
 os.environ["TEMP"] = tmp
 warnings.filterwarnings("ignore")
 torch.manual_seed(114514)
@@ -142,20 +142,22 @@ def forward_dml(ctx, x, scale):
 outside_index_root = os.getenv("outside_index_root")
 
 names = []
-for name in os.listdir(weight_root):
-    if name.endswith(".pth"):
-        names.append(name)
 index_paths = []
 
+def lookup_names(weight_root):
+    global names
+    for name in os.listdir(weight_root):
+        if name.endswith(".pth"):
+            names.append(name)
 
 def lookup_indices(index_root):
     global index_paths
-    for root, dirs, files in os.walk(index_root, topdown=False):
+    for root, _, files in os.walk(index_root, topdown=False):
         for name in files:
             if name.endswith(".index") and "trained" not in name:
-                index_paths.append("%s/%s" % (root, name))
-
+                index_paths.append(str(pathlib.Path(root, name)))
 
+lookup_names(weight_root)
 lookup_indices(index_root)
 lookup_indices(outside_index_root)
 uvr5_names = []
@@ -165,15 +167,12 @@ def lookup_indices(index_root):
 
 
 def change_choices():
+    global index_paths, names
     names = []
-    for name in os.listdir(weight_root):
-        if name.endswith(".pth"):
-            names.append(name)
+    lookup_names(weight_root)
     index_paths = []
-    for root, dirs, files in os.walk(index_root, topdown=False):
-        for name in files:
-            if name.endswith(".index") and "trained" not in name:
-                index_paths.append("%s/%s" % (root, name))
+    lookup_indices(index_root)
+    lookup_indices(outside_index_root)
     return {"choices": sorted(names), "__type__": "update"}, {
         "choices": sorted(index_paths),
         "__type__": "update",
@@ -223,16 +222,17 @@ def if_done_multi(done, ps):
 
 def preprocess_dataset(trainset_dir, exp_dir, sr, n_p):
     sr = sr_dict[sr]
-    os.makedirs("%s/logs/%s" % (now_dir, exp_dir), exist_ok=True)
-    f = open("%s/logs/%s/preprocess.log" % (now_dir, exp_dir), "w")
+    exp_path = pathlib.Path(now_dir, "logs", exp_dir)
+    os.makedirs(exp_path, exist_ok=True)
+    log_file_path = exp_path / "preprocess.log"
+    f = open(log_file_path, "w")
     f.close()
-    cmd = '"%s" infer/modules/train/preprocess.py "%s" %s %s "%s/logs/%s" %s %.1f' % (
+    cmd = '"%s" infer/modules/train/preprocess.py "%s" %s %s "%s" %s %.1f' % (
         config.python_cmd,
         trainset_dir,
         sr,
         n_p,
-        now_dir,
-        exp_dir,
+        str(exp_path),
         config.noparallel,
         config.preprocess_per,
     )
@@ -249,12 +249,12 @@ def preprocess_dataset(trainset_dir, exp_dir, sr, n_p):
         ),
     ).start()
     while 1:
-        with open("%s/logs/%s/preprocess.log" % (now_dir, exp_dir), "r") as f:
+        with open(log_file_path, "r") as f:
             yield (f.read())
         sleep(1)
         if done[0]:
             break
-    with open("%s/logs/%s/preprocess.log" % (now_dir, exp_dir), "r") as f:
+    with open(log_file_path, "r") as f:
         log = f.read()
     logger.info(log)
     yield log