songsenand
/
SUInput
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

添加日志记录和确保模型处于训练模式

This commit is contained in:
songsenand 2026-02-13 01:44:30 +08:00
parent 35e835f618
commit 982d0521d5
1 changed files with 6 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
src/trainer/model.py
View File

@ -7,6 +7,8 @@ from modelscope import AutoModel
import pickle import pickle
from importlib.resources import files from importlib.resources import files
from loguru import logger
from tqdm import tqdm from tqdm import tqdm
from .monitor import TrainingMonitor from .monitor import TrainingMonitor
@ -426,16 +428,20 @@ class MoEModel(nn.Module):
and global_step % eval_frequency == 0 and global_step % eval_frequency == 0
): ):
acc, _ = self.model_eval(eval_dataloader, criterion) acc, _ = self.model_eval(eval_dataloader, criterion)
super().train()
if monitor is not None: if monitor is not None:
monitor.add_step( monitor.add_step(
global_step, global_step,
{"loss": loss.item() * grad_accum_steps, "acc": acc}, {"loss": loss.item() * grad_accum_steps, "acc": acc},
) )
logger.info({"loss": loss.item() * grad_accum_steps, "acc": acc})
elif monitor is not None: elif monitor is not None:
# 仅记录训练损失 # 仅记录训练损失
monitor.add_step( monitor.add_step(
global_step, {"loss": loss.item() * grad_accum_steps} global_step, {"loss": loss.item() * grad_accum_steps}
) )
logger.info({"loss": loss.item() * grad_accum_steps})
# ============================ 使用示例 ============================ # ============================ 使用示例 ============================