Skip to content

SOTA Open Source TTS

Fix dataset and training loop

Lengyue01cf9bbafd4c
diff --git a/speech_lm/datasets/cultura_x.py b/speech_lm/datasets/cultura_x.py
index fa564ee..9445000 100644
--- a/speech_lm/datasets/cultura_x.py
+++ b/speech_lm/datasets/cultura_x.py
@@ -67,8 +67,8 @@ class CulturaXDataset(IterableDataset):
for filename in files:
try:
yield from self.parse_data(filename)
- except:
- log.exception(f"Failed to parse {filename}")
+ except Exception as e:
+ log.exception(f"Failed to parse {filename}: {e}")
def parse_data(self, filename: str):
url = f"https://huggingface.co/datasets/uonlp/CulturaX/resolve/main/{filename}"
diff --git a/speech_lm/train.py b/speech_lm/train.py
index ca82941..5b68f43 100644
--- a/speech_lm/train.py
+++ b/speech_lm/train.py
@@ -117,6 +117,9 @@ def main(cfg: DictConfig):
log.info(f"Optimizer: {optimizer}")
log.info(f"Scheduler: {scheduler}")
+ log.info(f"Setup fabric model & dataset")
+ model, optimizer, scheduler = fabric.setup(model, optimizer, scheduler)
+
# Build state
global_step = 0
@@ -141,9 +144,6 @@ def main(cfg: DictConfig):
global_step = remainder["global_step"]
log.info(f"Restored global step: {global_step}")
- log.info(f"Setup fabric model & dataset")
- model, optimizer, scheduler = fabric.setup(model, optimizer, scheduler)
-
train_dataloader = hydra.utils.instantiate(cfg.dataloader)
log.info(f"Dataloader: {train_dataloader}")