[ROCM] update paddle inference cmake, test=develop

[qili93]

PR types

New features

PR changes

Others

Describe

Update inference cmake file and add miopen cmake for paddle inference.

Output on MIOPEN 2.9

(base) root@h03r3n12:/opt python -c "import paddle; paddle.utils.run_check()"
Running verify PaddlePaddle program ...
W0601 08:14:24.456575   274 device_context.cc:430] Please NOTE: device: 0, GPU Compute Capability: 90.0, Driver API Version: 321.0, Runtime API Version: 3.1
W0601 08:14:24.460045   274 device_context.cc:443] device: 0, MIOpen Version: 2.9.0
W0601 08:14:24.747490   274 device_context.h:363] WARNING: device: 0. The installed Paddle is compiled with MIOPEN 2.11, but MIOPEN version in your machine is 2.9, which may cause serious incompatible bug. Please recompile or reinstall Paddle with compatible MIOPEN version.
PaddlePaddle works well on 1 GPU.
W0601 08:14:31.903820   274 fuse_all_reduce_op_pass.cc:76] Find all_reduce operators: 2. To make the speed faster, some all_reduce ops are fused during training, after fusion, the number of all_reduce ops is 2.
PaddlePaddle works well on 4 GPUs.
PaddlePaddle is installed successfully! Let's start deep learning with PaddlePaddle now.

Output on MIOPEN 2.11

(base) root@h03r3n12:/opt python -c "import paddle; paddle.utils.run_check()"
Running verify PaddlePaddle program ...
W0601 08:13:57.828878  2347 device_context.cc:430] Please NOTE: device: 0, GPU Compute Capability: 90.0, Driver API Version: 321.0, Runtime API Version: 3.1
W0601 08:13:57.832649  2347 device_context.cc:443] device: 0, MIOpen Version: 2.11.0
PaddlePaddle works well on 1 GPU.
W0601 08:14:05.520198  2347 fuse_all_reduce_op_pass.cc:76] Find all_reduce operators: 2. To make the speed faster, some all_reduce ops are fused during training, after fusion, the number of all_reduce ops is 2.
PaddlePaddle works well on 4 GPUs.
PaddlePaddle is installed successfully! Let's start deep learning with PaddlePaddle now.

Output of version.txt for Paddle Inference

GIT COMMIT ID: 0192b82f5bffcbd3a4b4a8737889270a78491498
WITH_MKL: ON
WITH_MKLDNN: ON
WITH_GPU: OFF
WITH_ROCM: ON
HIP version: 4.0.20496-4f163c68
MIOpen version: v2.11
CXX compiler version: 7.3.1

[paddle-bot-old]

Thanks for your contribution!
Please wait for the result of CI firstly. See Paddle CI Manual for details.

[wanghuancoder]

LGTM