Merge pull request IAHispano#692 from ShiromiyaG/delete-sliced

Option to delete sliced audio folders

core.py

@@ -468,6 +468,7 @@ def run_extract_script(
     sample_rate: int,
     embedder_model: str,
     embedder_model_custom: str = None,
+    delete_sliced_audios: bool = True,
 ):
 
     model_path = os.path.join(logs_path, model_name)
@@ -489,6 +490,7 @@ def run_extract_script(
                 sample_rate,
                 embedder_model,
                 embedder_model_custom,
+                delete_sliced_audios,
             ],
         ),
     ]
@@ -1329,6 +1331,14 @@ def parse_arguments():
         help=embedder_model_custom_description,
         default=None,
     )
+    extract_parser.add_argument(
+        "--delete_sliced_folders",
+        type=lambda x: bool(strtobool(x)),
+        choices=[True, False],
+        help="Auto delete sliced audio folders after embedder process.",
+        default=True,
+        required=False,
+    )
 
     # Parser for 'train' mode
     train_parser = subparsers.add_parser("train", help="Train an RVC model.")
@@ -1736,6 +1746,7 @@ def main():
                 sample_rate=args.sample_rate,
                 embedder_model=args.embedder_model,
                 embedder_model_custom=args.embedder_model_custom,
+                delete_sliced_folders=args.delete_sliced_folders,
             )
         elif args.mode == "train":
             run_train_script(

rvc/train/extract/extract.py

@@ -9,6 +9,8 @@
 import concurrent.futures
 import multiprocessing as mp
 import json
+import shutil
+from distutils.util import strtobool
 
 # Zluda
 if torch.cuda.is_available() and torch.cuda.get_device_name().endswith("[ZLUDA]"):
@@ -253,6 +255,7 @@ def run_embedding_extraction(
     sample_rate = sys.argv[8]
     embedder_model = sys.argv[9]
     embedder_model_custom = sys.argv[10] if len(sys.argv) > 10 else None
+    delete_sliced_audios = strtobool(sys.argv[11]) if len(sys.argv) > 11 else True
 
     # prep
     wav_path = os.path.join(exp_dir, "sliced_audios_16k")
@@ -295,3 +298,7 @@ def run_embedding_extraction(
     # Run Preparing Files
     generate_config(version, sample_rate, exp_dir)
     generate_filelist(pitch_guidance, exp_dir, version, sample_rate)
+
+    if delete_sliced_audios:
+        shutil.rmtree(os.path.join(exp_dir, "sliced_audios"), ignore_errors=True)
+        shutil.rmtree(os.path.join(exp_dir, "sliced_audios_16k"), ignore_errors=True)

tabs/train/train.py

@@ -489,14 +489,23 @@ def train_tab():
                         label="Upload .json", type="filepath", interactive=True
                     )
                 move_files_button = gr.Button("Move files to custom embedder folder")
-        pitch_guidance_extract = gr.Checkbox(
-            label=i18n("Pitch Guidance"),
-            info=i18n(
-                "By employing pitch guidance, it becomes feasible to mirror the intonation of the original voice, including its pitch. This feature is particularly valuable for singing and other scenarios where preserving the original melody or pitch pattern is essential."
-            ),
-            value=True,
-            interactive=True,
-        )
+        with gr.Row():
+            pitch_guidance_extract = gr.Checkbox(
+                label=i18n("Pitch Guidance"),
+                info=i18n(
+                    "By employing pitch guidance, it becomes feasible to mirror the intonation of the original voice, including its pitch. This feature is particularly valuable for singing and other scenarios where preserving the original melody or pitch pattern is essential."
+                ),
+                value=True,
+                interactive=True,
+            )
+            delete_sliced_audio = gr.Checkbox(
+                label=i18n("Delete Sliced Audio"),
+                info=i18n(
+                    "Delete the sliced audio folders after the embedder extraction process is completed."
+                ),
+                value=True,
+                interactive=True,
+            )
 
         with gr.Accordion(
             i18n(
@@ -556,6 +565,7 @@ def train_tab():
                 sampling_rate,
                 embedder_model,
                 embedder_model_custom,
+                delete_sliced_audio,
             ],
             outputs=[extract_output_info],
         )