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/configs/E2TTS_Base_train.yaml

@@ -41,4 +41,4 @@ ckpts:
   logger: wandb  # wandb | tensorboard | None
   save_per_updates: 50000  # save checkpoint per steps
   last_per_steps: 5000  # save last checkpoint per steps
-  save_dir: ckpts/${model.name}_${model.mel_spec.mel_spec_type}_${model.tokenizer}_${datasets.name}/${now:%Y-%m-%d}/${now:%H-%M-%S}
+  save_dir: ckpts/${model.name}_${model.mel_spec.mel_spec_type}_${model.tokenizer}_${datasets.name}

src/f5_tts/configs/E2TTS_Small_train.yaml

@@ -41,4 +41,4 @@ ckpts:
   logger: wandb  # wandb | tensorboard | None
   save_per_updates: 50000  # save checkpoint per steps
   last_per_steps: 5000  # save last checkpoint per steps
-  save_dir: ckpts/${model.name}_${model.mel_spec.mel_spec_type}_${model.tokenizer}_${datasets.name}/${now:%Y-%m-%d}/${now:%H-%M-%S}
+  save_dir: ckpts/${model.name}_${model.mel_spec.mel_spec_type}_${model.tokenizer}_${datasets.name}

src/f5_tts/configs/F5TTS_Base_train.yaml

@@ -43,4 +43,4 @@ ckpts:
   logger: wandb  # wandb | tensorboard | None
   save_per_updates: 50000  # save checkpoint per steps
   last_per_steps: 5000  # save last checkpoint per steps
-  save_dir: ckpts/${model.name}_${model.mel_spec.mel_spec_type}_${model.tokenizer}_${datasets.name}/${now:%Y-%m-%d}/${now:%H-%M-%S}
+  save_dir: ckpts/${model.name}_${model.mel_spec.mel_spec_type}_${model.tokenizer}_${datasets.name}

src/f5_tts/configs/F5TTS_Small_train.yaml

@@ -43,4 +43,4 @@ ckpts:
   logger: wandb  # wandb | tensorboard | None
   save_per_updates: 50000  # save checkpoint per steps
   last_per_steps: 5000  # save last checkpoint per steps
-  save_dir: ckpts/${model.name}_${model.mel_spec.mel_spec_type}_${model.tokenizer}_${datasets.name}/${now:%Y-%m-%d}/${now:%H-%M-%S}
+  save_dir: ckpts/${model.name}_${model.mel_spec.mel_spec_type}_${model.tokenizer}_${datasets.name}

src/f5_tts/infer/utils_infer.py

@@ -138,7 +138,11 @@ asr_pipe = None
 def initialize_asr_pipeline(device: str = device, dtype=None):
     if dtype is None:
         dtype = (
-            torch.float16 if "cuda" in device and torch.cuda.get_device_properties(device).major >= 6 else torch.float32
+            torch.float16
+            if "cuda" in device
+            and torch.cuda.get_device_properties(device).major >= 6
+            and not torch.cuda.get_device_name().endswith("[ZLUDA]")
+            else torch.float32
         )
     global asr_pipe
     asr_pipe = pipeline(
@@ -171,7 +175,11 @@ def transcribe(ref_audio, language=None):
 def load_checkpoint(model, ckpt_path, device: str, dtype=None, use_ema=True):
     if dtype is None:
         dtype = (
-            torch.float16 if "cuda" in device and torch.cuda.get_device_properties(device).major >= 6 else torch.float32
+            torch.float16
+            if "cuda" in device
+            and torch.cuda.get_device_properties(device).major >= 6
+            and not torch.cuda.get_device_name().endswith("[ZLUDA]")
+            else torch.float32
         )
     model = model.to(dtype)