Hi all:
I tried to run OpenVINO benchmark_app on my AWS EC2 instance but it failed running with option "-infer_precision CPU:bf16"
AWS instance:
- Type: m7i-flex large
- OS: Windows Server 2019 (AMI ID: ami-065b889ab5c33720e )
- CPU: Intel Xeon Platinum 8488C, 1 core, 2 threads (Code name: Sapphire Rapids)
- Memory: 8G
- HDD: 30G
(I thought this CPU should support precision FP16 & FP32)
Model:
- asl-recognition-0004
- downloaded by "omz_downloader --name asl-recognition-0004 --precisions FP16 --output_dir myLocalPath"
What I did:
- (create and activate python 3.8 environment)
- pip install openvino==2022.3
- pip install openvino-dev==2022.3
- omz_downloader --name asl-recognition-0004 --precisions FP16 --output_dir myLocalPath
- benchmark_app -infer_precision CPU:bf16 -m myLocalPath\asl-recognition-0004.xml
- (then it failed)
Although precision of the model is FP16, I tried to set -infer_precision to CPU:f32, then it ran successfully.
I also downloaded asl-recognition-0004 FP32 ver. and benchmark_app still fails when running with "-infer_precision CPU:bf16"
It now seems benchmark_app always failed when running with option "-infer_precision CPU:bf16", no matter which precision of input model.
Sorry but I failed to find related documents about this case.
Best Regards
Clin B
Link Copied
Hi Clin B,
I'm able to run both FP16 and FP32 infer_precision with Benchmark_App with the latest OpenVINO 2023.0 version. Please refer below:
1) benchmark_app -infer_precision CPU:bf16 -m \<Download_dir>\FP16\asl-recognition-0004.xml
2) benchmark_app -infer_precision CPU:f32 -m \<Download_dir>\FP32\asl-recognition-0004.xml
Regards,
Aznie
