docker 环境：
paddle-bfloat 0.1.7，paddleocr 2.7.0.3，paddlepaddle 2.5.1

执行表格识别实战的案例，执行如下命令：

python tools/train.py -c configs/table/SLANet_ch.yml -o \
Global.pretrained_model=./pretrain_models/ch_ppstructure_mobile_v2.0_SLANet_train/best_accuracy.pdparams \
Global.save_model_dir=output/SLANet_ch/ \
Global.eval_batch_step="[0, 375]" \
Optimizer.lr.name=Const \
Optimizer.lr.learning_rate=0.0005 \
Optimizer.lr.warmup_epoch=0 \
Train.dataset.data_dir=/home/paddle/data/data165849 \
Train.dataset.label_file_list=[/home/paddle/train.txt] \
Eval.dataset.data_dir=/home/paddle/data/data165849 \
Eval.dataset.label_file_list=[/home/paddle/val.txt]

上面yml文件里的use_gpu设置为false

出现如下错误提示：

File "/opt/conda/lib/python3.8/site-packages/paddle/nn/layer/norm.py", line 1547, in forward
sync_batch_norm_out, _, _, _, _, _ = _C_ops.sync_batch_norm_(
RuntimeError: (NotFound) The kernel `sync_batch_norm` is not registered.
[Hint: Expected iter != kernels_.end(), but received iter == kernels_.end().] (at /paddle/paddle/phi/core/kernel_factory.cc:174)