Spaces:
Running
on
Zero
Running
on
Zero
mrfakename
commited on
Commit
•
490128a
1
Parent(s):
910037c
Sync from GitHub repo
Browse filesThis 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
CHANGED
@@ -41,4 +41,4 @@ ckpts:
|
|
41 |
logger: wandb # wandb | tensorboard | None
|
42 |
save_per_updates: 50000 # save checkpoint per steps
|
43 |
last_per_steps: 5000 # save last checkpoint per steps
|
44 |
-
save_dir: ckpts/${model.name}_${model.mel_spec.mel_spec_type}_${model.tokenizer}_${datasets.name}
|
|
|
41 |
logger: wandb # wandb | tensorboard | None
|
42 |
save_per_updates: 50000 # save checkpoint per steps
|
43 |
last_per_steps: 5000 # save last checkpoint per steps
|
44 |
+
save_dir: ckpts/${model.name}_${model.mel_spec.mel_spec_type}_${model.tokenizer}_${datasets.name}
|
src/f5_tts/configs/E2TTS_Small_train.yaml
CHANGED
@@ -41,4 +41,4 @@ ckpts:
|
|
41 |
logger: wandb # wandb | tensorboard | None
|
42 |
save_per_updates: 50000 # save checkpoint per steps
|
43 |
last_per_steps: 5000 # save last checkpoint per steps
|
44 |
-
save_dir: ckpts/${model.name}_${model.mel_spec.mel_spec_type}_${model.tokenizer}_${datasets.name}
|
|
|
41 |
logger: wandb # wandb | tensorboard | None
|
42 |
save_per_updates: 50000 # save checkpoint per steps
|
43 |
last_per_steps: 5000 # save last checkpoint per steps
|
44 |
+
save_dir: ckpts/${model.name}_${model.mel_spec.mel_spec_type}_${model.tokenizer}_${datasets.name}
|
src/f5_tts/configs/F5TTS_Base_train.yaml
CHANGED
@@ -43,4 +43,4 @@ ckpts:
|
|
43 |
logger: wandb # wandb | tensorboard | None
|
44 |
save_per_updates: 50000 # save checkpoint per steps
|
45 |
last_per_steps: 5000 # save last checkpoint per steps
|
46 |
-
save_dir: ckpts/${model.name}_${model.mel_spec.mel_spec_type}_${model.tokenizer}_${datasets.name}
|
|
|
43 |
logger: wandb # wandb | tensorboard | None
|
44 |
save_per_updates: 50000 # save checkpoint per steps
|
45 |
last_per_steps: 5000 # save last checkpoint per steps
|
46 |
+
save_dir: ckpts/${model.name}_${model.mel_spec.mel_spec_type}_${model.tokenizer}_${datasets.name}
|
src/f5_tts/configs/F5TTS_Small_train.yaml
CHANGED
@@ -43,4 +43,4 @@ ckpts:
|
|
43 |
logger: wandb # wandb | tensorboard | None
|
44 |
save_per_updates: 50000 # save checkpoint per steps
|
45 |
last_per_steps: 5000 # save last checkpoint per steps
|
46 |
-
save_dir: ckpts/${model.name}_${model.mel_spec.mel_spec_type}_${model.tokenizer}_${datasets.name}
|
|
|
43 |
logger: wandb # wandb | tensorboard | None
|
44 |
save_per_updates: 50000 # save checkpoint per steps
|
45 |
last_per_steps: 5000 # save last checkpoint per steps
|
46 |
+
save_dir: ckpts/${model.name}_${model.mel_spec.mel_spec_type}_${model.tokenizer}_${datasets.name}
|
src/f5_tts/infer/utils_infer.py
CHANGED
@@ -138,7 +138,11 @@ asr_pipe = None
|
|
138 |
def initialize_asr_pipeline(device: str = device, dtype=None):
|
139 |
if dtype is None:
|
140 |
dtype = (
|
141 |
-
torch.float16
|
|
|
|
|
|
|
|
|
142 |
)
|
143 |
global asr_pipe
|
144 |
asr_pipe = pipeline(
|
@@ -171,7 +175,11 @@ def transcribe(ref_audio, language=None):
|
|
171 |
def load_checkpoint(model, ckpt_path, device: str, dtype=None, use_ema=True):
|
172 |
if dtype is None:
|
173 |
dtype = (
|
174 |
-
torch.float16
|
|
|
|
|
|
|
|
|
175 |
)
|
176 |
model = model.to(dtype)
|
177 |
|
|
|
138 |
def initialize_asr_pipeline(device: str = device, dtype=None):
|
139 |
if dtype is None:
|
140 |
dtype = (
|
141 |
+
torch.float16
|
142 |
+
if "cuda" in device
|
143 |
+
and torch.cuda.get_device_properties(device).major >= 6
|
144 |
+
and not torch.cuda.get_device_name().endswith("[ZLUDA]")
|
145 |
+
else torch.float32
|
146 |
)
|
147 |
global asr_pipe
|
148 |
asr_pipe = pipeline(
|
|
|
175 |
def load_checkpoint(model, ckpt_path, device: str, dtype=None, use_ema=True):
|
176 |
if dtype is None:
|
177 |
dtype = (
|
178 |
+
torch.float16
|
179 |
+
if "cuda" in device
|
180 |
+
and torch.cuda.get_device_properties(device).major >= 6
|
181 |
+
and not torch.cuda.get_device_name().endswith("[ZLUDA]")
|
182 |
+
else torch.float32
|
183 |
)
|
184 |
model = model.to(dtype)
|
185 |
|