From 6ca0b838d48106030984eacf204e8f1f2f05985b Mon Sep 17 00:00:00 2001
From: 游雁 <zhifu.gzf@alibaba-inc.com>
Date: 星期四, 13 六月 2024 16:07:49 +0800
Subject: [PATCH] decoding
---
funasr/train_utils/trainer_ds.py | 8 +++++---
1 files changed, 5 insertions(+), 3 deletions(-)
diff --git a/funasr/train_utils/trainer_ds.py b/funasr/train_utils/trainer_ds.py
index 22be426..e6b6284 100644
--- a/funasr/train_utils/trainer_ds.py
+++ b/funasr/train_utils/trainer_ds.py
@@ -78,7 +78,7 @@
output_dir (str): The directory where model checkpoints will be saved. Default is './'.
resume (str, optional): The file path to a checkpoint to resume training from.
"""
- self.rank = kwargs.get("rank", 0)
+ self.rank = rank
self.local_rank = local_rank
self.world_size = world_size
self.use_ddp = use_ddp
@@ -287,7 +287,9 @@
elif self.use_fsdp:
pass
elif self.rank == 0:
- logging.info(f"Save checkpoint: {epoch}, rank: {self.local_rank}\n")
+ logging.info(
+ f"Save checkpoint: {epoch}, rank: {self.rank}, local_rank: {self.local_rank}\n"
+ )
# self.step_or_epoch += 1
state = {
"epoch": epoch,
@@ -319,7 +321,7 @@
for k_ex in self.effective_save_name_excludes:
k_tmp = k.replace("module.", "")
if k.startswith(k_ex):
- logging.info(f"key: {{k}} matching: {k_ex}, not save it")
+ logging.info(f"key: {k} matching: {k_ex}, not save it")
break
else:
dst_state_dict[k] = state_dict[k]
--
Gitblit v1.9.1