speech_recognition/FunASR
1
0
Fork 0
mirror of https://github.com/modelscope/FunASR synced 2025-09-15 14:48:36 +08:00
Code Issues Actions Packages Projects Releases Wiki Activity

update

Browse Source
This commit is contained in:
游雁 2024-03-22 21:08:55 +08:00
parent 334422ead3
commit 6ce4909ecc
1 changed files with 14 additions and 14 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

28
funasr/train_utils/trainer.py
View File

@ -290,13 +290,13 @@ class Trainer:
self.train_loss_avg = (self.train_loss_avg*batch_idx + loss.detach().cpu().item())/(batch_idx+1)
if "acc" in stats:
self.train_acc_avg = (self.train_acc_avg * batch_idx + stats["acc"].detach().cpu().item()) / (batch_idx + 1)
if self.use_ddp or self.use_fsdp:
train_loss_avg = torch.tensor(self.train_loss_avg, dtype=torch.float32).to(self.device)
train_acc_avg = torch.tensor(self.train_acc_avg, dtype=torch.float32).to(self.device)
dist.all_reduce(train_loss_avg, op=dist.ReduceOp.SUM)
dist.all_reduce(train_acc_avg, op=dist.ReduceOp.SUM)
self.train_loss_avg = train_loss_avg.detach().cpu().item() / self.world_size
self.train_acc_avg = train_acc_avg.detach().cpu().item() / self.world_size
# if self.use_ddp or self.use_fsdp:
# train_loss_avg = torch.tensor(self.train_loss_avg, dtype=torch.float32).to(self.device)
# train_acc_avg = torch.tensor(self.train_acc_avg, dtype=torch.float32).to(self.device)
# dist.all_reduce(train_loss_avg, op=dist.ReduceOp.SUM)
# dist.all_reduce(train_acc_avg, op=dist.ReduceOp.SUM)
# self.train_loss_avg = train_loss_avg.detach().cpu().item() / self.world_size
# self.train_acc_avg = train_acc_avg.detach().cpu().item() / self.world_size
# Perform an optimizer step only after accumulating enough gradients
@ -412,13 +412,13 @@ class Trainer:
self.val_loss_avg = (self.val_loss_avg*batch_idx + loss.detach().cpu().item())/(batch_idx+1)
if "acc" in stats:
self.val_acc_avg = (self.val_acc_avg * batch_idx + stats["acc"].detach().cpu().item()) / (batch_idx + 1)
if self.use_ddp or self.use_fsdp:
val_loss_avg = torch.tensor(self.val_loss_avg, dtype=torch.float32).to(self.device)
val_acc_avg = torch.tensor(self.val_acc_avg, dtype=torch.float32).to(self.device)
dist.all_reduce(val_loss_avg, op=dist.ReduceOp.SUM)
dist.all_reduce(val_acc_avg, op=dist.ReduceOp.SUM)
self.val_loss_avg = val_loss_avg.detach().cpu().item() / self.world_size
self.val_acc_avg = val_acc_avg.detach().cpu().item() / self.world_size
# if self.use_ddp or self.use_fsdp:
# val_loss_avg = torch.tensor(self.val_loss_avg, dtype=torch.float32).to(self.device)
# val_acc_avg = torch.tensor(self.val_acc_avg, dtype=torch.float32).to(self.device)
# dist.all_reduce(val_loss_avg, op=dist.ReduceOp.SUM)
# dist.all_reduce(val_acc_avg, op=dist.ReduceOp.SUM)
# self.val_loss_avg = val_loss_avg.detach().cpu().item() / self.world_size
# self.val_acc_avg = val_acc_avg.detach().cpu().item() / self.world_size
batch_num_epoch = -1
if hasattr(dataloader_val, "__len__"):