From 9ba0dbd98bf69c830dfcfde8f109a400cb65e4e5 Mon Sep 17 00:00:00 2001
From: 雾聪 <wucong.lyb@alibaba-inc.com>
Date: 星期五, 29 三月 2024 17:24:59 +0800
Subject: [PATCH] fix func Forward
---
funasr/train_utils/trainer.py | 11 ++++++++---
1 files changed, 8 insertions(+), 3 deletions(-)
diff --git a/funasr/train_utils/trainer.py b/funasr/train_utils/trainer.py
index 5b280bf..723a149 100644
--- a/funasr/train_utils/trainer.py
+++ b/funasr/train_utils/trainer.py
@@ -14,6 +14,7 @@
from funasr.train_utils.device_funcs import to_device
from funasr.train_utils.recursive_op import recursive_average
from funasr.train_utils.average_nbest_models import average_checkpoints
+from torch.distributed.fsdp.sharded_grad_scaler import ShardedGradScaler
@contextmanager
def maybe_autocast(enabled):
@@ -84,7 +85,9 @@
self.batch_total = 0
self.use_fp16 = use_fp16
self.disable_gpu_cache = kwargs.get("disable_gpu_cache", True)
- self.scaler = GradScaler(enabled=use_fp16) if use_fp16 else None
+ scaler = GradScaler(enabled=use_fp16) if use_fp16 else None
+ scaler = ShardedGradScaler(enabled=use_fp16) if use_ddp else scaler
+ self.scaler = scaler
try:
@@ -160,7 +163,7 @@
self.scaler.load_state_dict(checkpoint['scaler_state'])
print(f"Checkpoint loaded successfully from '{ckpt}'")
else:
- print(f"No checkpoint found at '{ckpt}', starting from scratch")
+ print(f"No checkpoint found at '{ckpt}', does not resume status!")
if self.use_ddp or self.use_fsdp:
dist.barrier()
@@ -398,4 +401,6 @@
epoch * len(self.dataloader_val) + batch_idx)
for key, var in speed_stats.items():
self.writer.add_scalar(f'rank{self.local_rank}_{key}/val', eval(var),
- epoch * len(self.dataloader_val) + batch_idx)
\ No newline at end of file
+ epoch * len(self.dataloader_val) + batch_idx)
+
+ self.model.train()
\ No newline at end of file
--
Gitblit v1.9.1