inference by link

app/infer.py

@@ -12,23 +12,20 @@ from .zero import zero
 from .model import device
 import yt_dlp
 
-
 def download_audio(url):
     ydl_opts = {
-        "format": "bestaudio/best",
-        "outtmpl": "ytdl/%(title)s.%(ext)s",
-        "postprocessors": [
-            {
-                "key": "FFmpegExtractAudio",
-                "preferredcodec": "wav",
-                "preferredquality": "192",
-            }
-        ],
+        'format': 'bestaudio/best',
+        'outtmpl': 'ytdl/%(title)s.%(ext)s',
+        'postprocessors': [{
+            'key': 'FFmpegExtractAudio',
+            'preferredcodec': 'wav',
+            'preferredquality': '192',
+        }],
     }
 
     with yt_dlp.YoutubeDL(ydl_opts) as ydl:
         info_dict = ydl.extract_info(url, download=True)
-        file_path = ydl.prepare_filename(info_dict).rsplit(".", 1)[0] + ".wav"
+        file_path = ydl.prepare_filename(info_dict).rsplit('.', 1)[0] + '.wav'
         sample_rate, audio_data = read(file_path)
         audio_array = np.asarray(audio_data, dtype=np.int16)
 
@@ -49,6 +46,7 @@ def infer(
     ext = Path(original_audio).suffix
     original_audio_hashed = os.path.join(exp_dir, f"{original_audio_hash}{ext}")
     shutil.copy(original_audio, original_audio_hashed)
+    
 
     out = os.path.join("separated", "htdemucs", original_audio_hash, "vocals.wav")
     if not os.path.exists(out):
@@ -104,22 +102,18 @@ class InferenceTab:
                 type="filepath",
                 show_download_button=True,
             )
-            with gr.Accordion("inference by Link", open=False):
+            with gr.Accordion("inference by Link",open=False):
                 with gr.Row():
                     youtube_link = gr.Textbox(
-                        label="Link",
-                        placeholder="Paste the link here",
-                        interactive=True,
-                    )
+                        label = "Link",
+                        placeholder = "Paste the link here",
+                        interactive = True
+                )
                 with gr.Row():
-                    gr.Markdown(
-                        "You can paste the link to the video/audio from many sites, check the complete list [here](https://github.com/yt-dlp/yt-dlp/blob/master/supportedsites.md)"
-                    )
+                    gr.Markdown("You can paste the link to the video/audio from many sites, check the complete list [here](https://github.com/yt-dlp/yt-dlp/blob/master/supportedsites.md)")
                 with gr.Row():
-                    download_button = gr.Button("Download!", variant="primary")
-                    download_button.click(
-                        download_audio, [youtube_link], [self.original_audio]
-                    )
+                    download_button = gr.Button("Download!",variant = "primary")
+                    download_button.click(download_audio, [youtube_link], [self.original_audio])
 
             with gr.Column():
                 self.pitch_mod = gr.Slider(

pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "zerorvc"
-version = "0.0.10"
+version = "0.0.8"
 authors = [{ name = "Jacob Lin", email = "[email protected]" }]
 description = "Run Retrieval-based Voice Conversion training and inference with ease."
 readme = "README.md"

requirements.txt

@@ -1,4 +1,4 @@
-zerorvc>=0.0.10
+zerorvc>=0.0.8
 
 # gradio app deps
 gradio==4.37.2

zerorvc/preprocess/preprocess.py

@@ -19,7 +19,7 @@ class Preprocessor:
         self.sr = sr
         self.bh, self.ah = signal.butter(N=5, Wn=48, btype="high", fs=self.sr)
         self.max_slice_length = max_slice_length
-        self.min_slice_length = min_slice_length
+        self.max_slice_length = min_slice_length
         self.overlap = 0.3
         self.tail = self.max_slice_length + self.overlap
         self.max = 0.9

zerorvc/trainer.py

@@ -9,7 +9,6 @@ from torch.utils.data import DataLoader
 from huggingface_hub import HfApi, upload_folder
 from accelerate import Accelerator
 from datasets import Dataset
-from .pretrained import pretrained_checkpoints
 from .constants import *
 
 
@@ -101,15 +100,15 @@ class RVCTrainer:
         self.checkpoint_dir = checkpoint_dir
         self.sr = sr
 
-    def latest_checkpoint(self, fallback_to_pretrained: bool = True):
+    def latest_checkpoint(self):
         files_g = glob(os.path.join(self.checkpoint_dir, "G_*.pth"))
         if not files_g:
-            return pretrained_checkpoints() if fallback_to_pretrained else None
+            return None
         latest_g = max(files_g, key=os.path.getctime)
 
         files_d = glob(os.path.join(self.checkpoint_dir, "D_*.pth"))
         if not files_d:
-            return pretrained_checkpoints() if fallback_to_pretrained else None
+            return None
         latest_d = max(files_d, key=os.path.getctime)
 
         return latest_g, latest_d