未验证 提交 5eafd467 编写于 作者: D Double_V 提交者: GitHub

Merge pull request #2230 from LDOUBLEV/fix_dyg

enable memory optim
...@@ -128,7 +128,8 @@ def create_predictor(args, mode, logger): ...@@ -128,7 +128,8 @@ def create_predictor(args, mode, logger):
#config.set_mkldnn_op({'conv2d', 'depthwise_conv2d', 'pool2d', 'batch_norm'}) #config.set_mkldnn_op({'conv2d', 'depthwise_conv2d', 'pool2d', 'batch_norm'})
args.rec_batch_num = 1 args.rec_batch_num = 1
# config.enable_memory_optim() # enable memory optim
config.enable_memory_optim()
config.disable_glog_info() config.disable_glog_info()
config.delete_pass("conv_transpose_eltwiseadd_bn_fuse_pass") config.delete_pass("conv_transpose_eltwiseadd_bn_fuse_pass")
......
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册