From 28ccfbfc51068a663a80764e14074df5edf2b5ba Mon Sep 17 00:00:00 2001
From: kongdeqiang <kongdeqiang960204@163.com>
Date: 星期五, 13 三月 2026 17:41:41 +0800
Subject: [PATCH] 提交
---
funasr/bin/train.py | 21 +++++++++++++++++----
1 files changed, 17 insertions(+), 4 deletions(-)
diff --git a/funasr/bin/train.py b/funasr/bin/train.py
index 643df71..c56d047 100644
--- a/funasr/bin/train.py
+++ b/funasr/bin/train.py
@@ -20,6 +20,7 @@
from torch.distributed.fsdp import FullyShardedDataParallel as FSDP
from torch.distributed.algorithms.join import Join
from torch.distributed.fsdp.sharded_grad_scaler import ShardedGradScaler
+from tensorboardX import SummaryWriter
from funasr.train_utils.average_nbest_models import average_checkpoints
from funasr.register import tables
@@ -27,7 +28,7 @@
from funasr.train_utils.trainer import Trainer
from funasr.schedulers import scheduler_classes
from funasr.train_utils.initialize import initialize
-from funasr.download.download_from_hub import download_model
+from funasr.download.download_model_from_hub import download_model
from funasr.models.lora.utils import mark_only_lora_as_trainable
from funasr.train_utils.set_all_random_seed import set_all_random_seed
from funasr.train_utils.load_pretrained_model import load_pretrained_model
@@ -191,8 +192,6 @@
tensorboard_dir = os.path.join(kwargs.get("output_dir"), "tensorboard")
os.makedirs(tensorboard_dir, exist_ok=True)
try:
- from tensorboardX import SummaryWriter
-
writer = SummaryWriter(tensorboard_dir) # if trainer.rank == 0 else None
except:
writer = None
@@ -202,6 +201,7 @@
time1 = time.perf_counter()
for data_split_i in range(trainer.start_data_split_i, dataloader.data_split_num):
+ time_slice_i = time.perf_counter()
dataloader_tr, dataloader_val = dataloader.build_iter(
epoch, data_split_i=data_split_i, start_step=trainer.start_step
)
@@ -221,7 +221,18 @@
)
trainer.start_step = 0
- torch.cuda.empty_cache()
+ device = next(model.parameters()).device
+ if device.type == "cuda":
+ with torch.cuda.device(device):
+ torch.cuda.empty_cache()
+
+ time_escaped = (time.perf_counter() - time_slice_i) / 3600.0
+ logging.info(
+ f"rank: {local_rank}, "
+ f"time_escaped_epoch: {time_escaped:.3f} hours, "
+ f"estimated to finish {dataloader.data_split_num} data_slices, remaining: {dataloader.data_split_num-data_split_i} slices, {(dataloader.data_split_num-data_split_i)*time_escaped:.3f} hours, "
+ f"epoch: {trainer.max_epoch - epoch} epochs, {((trainer.max_epoch - epoch - 1)*dataloader.data_split_num + dataloader.data_split_num-data_split_i)*time_escaped:.3f} hours\n"
+ )
trainer.start_data_split_i = 0
trainer.validate_epoch(
@@ -241,6 +252,8 @@
f"estimated to finish {trainer.max_epoch} "
f"epoch: {(trainer.max_epoch - epoch) * time_escaped:.3f} hours\n"
)
+ trainer.train_acc_avg = 0.0
+ trainer.train_loss_avg = 0.0
if trainer.rank == 0:
average_checkpoints(trainer.output_dir, trainer.avg_nbest_model)
--
Gitblit v1.9.1