fix(train_page): 捕获训练过程中的异常并终止 TensorBoard 进程

- 在训练过程中添加异常捕获,将异常信息转换为 gr.Error 抛出
- 确保在发生异常时也能终止 TensorBoard 子进程
This commit is contained in:
carry 2025-04-20 21:40:46 +08:00
parent 0a4efa5641
commit 9806334517

View File

@ -82,6 +82,8 @@ def train_page():
dataset, new_training_dir,
learning_rate, per_device_train_batch_size, epoch,
save_steps, lora_rank)
except Exception as e:
raise gr.Error(str(e))
finally:
# 确保训练结束后终止 TensorBoard 子进程
tensorboard_process.terminate()