Skip to content

Open-source speech recognition toolkit for training, inference, streaming ASR, VAD, punctuation, speaker diarization pipelines, and OpenAI-compatible/MCP serving.

finetune

游雁a70f5b3edf22
diff --git a/funasr/bin/train.py b/funasr/bin/train.py
index 5cf54daa2..e446e5404 100644
--- a/funasr/bin/train.py
+++ b/funasr/bin/train.py
@@ -173,10 +173,10 @@ def main(**kwargs):
except:
writer = None
- # if use_ddp or use_fsdp:
- # context = Join([model])
- # else:
- context = nullcontext()
+ if use_ddp or use_fsdp:
+ context = Join([model])
+ else:
+ context = nullcontext()
for epoch in range(trainer.start_epoch, trainer.max_epoch + 1):
time1 = time.perf_counter()
@@ -192,13 +192,14 @@ def main(**kwargs):
epoch=epoch,
writer=writer
)
+ with context:
+ trainer.validate_epoch(
+ model=model,
+ dataloader_val=dataloader_val,
+ epoch=epoch,
+ writer=writer
+ )
scheduler.step()
- trainer.validate_epoch(
- model=model,
- dataloader_val=dataloader_val,
- epoch=epoch,
- writer=writer
- )
trainer.save_checkpoint(epoch, model=model, optim=optim, scheduler=scheduler, scaler=scaler)
diff --git a/funasr/train_utils/trainer.py b/funasr/train_utils/trainer.py
index cf2348314..e554aca6d 100644
--- a/funasr/train_utils/trainer.py
+++ b/funasr/train_utils/trainer.py
@@ -398,7 +398,7 @@ class Trainer:
speed_stats = {}
time5 = time.perf_counter()
# iterator_stop = torch.tensor(0).to(self.device)
-
+ dataloader_val.batch_sampler.set_epoch(epoch)
for batch_idx, batch in enumerate(dataloader_val):
# if self.use_ddp or self.use_fsdp:
# dist.all_reduce(iterator_stop, dist.ReduceOp.SUM)