Sync from GitHub repo

This Space is synced from the GitHub repo: https://github.com/SWivid/F5-TTS. Please submit contributions to the Space there

src/f5_tts/api.py

@@ -32,6 +32,7 @@ class F5TTS:
         vocoder_name="vocos",
         local_path=None,
         device=None,
+        hf_cache_dir=None,
     ):
         # Initialize parameters
         self.final_wave = None
@@ -46,29 +47,31 @@ class F5TTS:
         )
 
         # Load models
-        self.load_vocoder_model(vocoder_name, local_path=local_path)
-        self.load_ema_model(model_type, ckpt_file, vocoder_name, vocab_file, ode_method, use_ema, local_path=local_path)
+        self.load_vocoder_model(vocoder_name, local_path=local_path, hf_cache_dir=hf_cache_dir)
+        self.load_ema_model(
+            model_type, ckpt_file, vocoder_name, vocab_file, ode_method, use_ema, hf_cache_dir=hf_cache_dir
+        )
 
-    def load_vocoder_model(self, vocoder_name, local_path=None):
-        self.vocoder = load_vocoder(vocoder_name, local_path is not None, local_path, self.device)
+    def load_vocoder_model(self, vocoder_name, local_path=None, hf_cache_dir=None):
+        self.vocoder = load_vocoder(vocoder_name, local_path is not None, local_path, self.device, hf_cache_dir)
 
-    def load_ema_model(self, model_type, ckpt_file, mel_spec_type, vocab_file, ode_method, use_ema, local_path=None):
+    def load_ema_model(self, model_type, ckpt_file, mel_spec_type, vocab_file, ode_method, use_ema, hf_cache_dir=None):
         if model_type == "F5-TTS":
             if not ckpt_file:
                 if mel_spec_type == "vocos":
                     ckpt_file = str(
-                        cached_path("hf://SWivid/F5-TTS/F5TTS_Base/model_1200000.safetensors", cache_dir=local_path)
+                        cached_path("hf://SWivid/F5-TTS/F5TTS_Base/model_1200000.safetensors", cache_dir=hf_cache_dir)
                     )
                 elif mel_spec_type == "bigvgan":
                     ckpt_file = str(
-                        cached_path("hf://SWivid/F5-TTS/F5TTS_Base_bigvgan/model_1250000.pt", cache_dir=local_path)
+                        cached_path("hf://SWivid/F5-TTS/F5TTS_Base_bigvgan/model_1250000.pt", cache_dir=hf_cache_dir)
                     )
             model_cfg = dict(dim=1024, depth=22, heads=16, ff_mult=2, text_dim=512, conv_layers=4)
             model_cls = DiT
         elif model_type == "E2-TTS":
             if not ckpt_file:
                 ckpt_file = str(
-                    cached_path("hf://SWivid/E2-TTS/E2TTS_Base/model_1200000.safetensors", cache_dir=local_path)
+                    cached_path("hf://SWivid/E2-TTS/E2TTS_Base/model_1200000.safetensors", cache_dir=hf_cache_dir)
                 )
             model_cfg = dict(dim=1024, depth=24, heads=16, ff_mult=4)
             model_cls = UNetT

src/f5_tts/infer/utils_infer.py

@@ -90,18 +90,18 @@ def chunk_text(text, max_chars=135):
 
 
 # load vocoder
-def load_vocoder(vocoder_name="vocos", is_local=False, local_path=None, device=device):
+def load_vocoder(vocoder_name="vocos", is_local=False, local_path="", device=device, hf_cache_dir=None):
     if vocoder_name == "vocos":
         # vocoder = Vocos.from_pretrained("charactr/vocos-mel-24khz").to(device)
-        if is_local and local_path is not None:
+        if is_local:
             print(f"Load vocos from local path {local_path}")
             config_path = f"{local_path}/config.yaml"
             model_path = f"{local_path}/pytorch_model.bin"
         else:
             print("Download Vocos from huggingface charactr/vocos-mel-24khz")
             repo_id = "charactr/vocos-mel-24khz"
-            config_path = hf_hub_download(repo_id=repo_id, cache_dir=local_path, filename="config.yaml")
-            model_path = hf_hub_download(repo_id=repo_id, cache_dir=local_path, filename="pytorch_model.bin")
+            config_path = hf_hub_download(repo_id=repo_id, cache_dir=hf_cache_dir, filename="config.yaml")
+            model_path = hf_hub_download(repo_id=repo_id, cache_dir=hf_cache_dir, filename="pytorch_model.bin")
         vocoder = Vocos.from_hparams(config_path)
         state_dict = torch.load(model_path, map_location="cpu", weights_only=True)
         from vocos.feature_extractors import EncodecFeatures
@@ -119,11 +119,11 @@ def load_vocoder(vocoder_name="vocos", is_local=False, local_path=None, device=d
             from third_party.BigVGAN import bigvgan
         except ImportError:
             print("You need to follow the README to init submodule and change the BigVGAN source code.")
-        if is_local and local_path is not None:
+        if is_local:
             """download from https://huggingface.co/nvidia/bigvgan_v2_24khz_100band_256x/tree/main"""
             vocoder = bigvgan.BigVGAN.from_pretrained(local_path, use_cuda_kernel=False)
         else:
-            local_path = snapshot_download(repo_id="nvidia/bigvgan_v2_24khz_100band_256x", cache_dir=local_path)
+            local_path = snapshot_download(repo_id="nvidia/bigvgan_v2_24khz_100band_256x", cache_dir=hf_cache_dir)
             vocoder = bigvgan.BigVGAN.from_pretrained(local_path, use_cuda_kernel=False)
 
         vocoder.remove_weight_norm()