Created by: FrostML
Recently, there are some passes failed during the inference. Hence, it's necessary to have a test in fluid to check whether passes still work. This PR pays attention to passes' precision. And add some unit tests.
- A test for checking seqconv_eltadd_relu_fuse_pass' precision is added.
- A test for checking conv_bn_fuse_pass' precision is added.
- A test for checking conv_eltadd_act_fuse_pass' precision is added.
- A test for checking conv_eltadd_fuse_pass' precision is added.
- A test for checking conv_eltadd2_act_fuse_pass' precision is added.
- A test for checking fc_fuse_pass' precision is added.
- A test for checking repeated_fc_relu_fuse_pass' precision is added.