How to confirm if the current hardware supports operators such as fp16/int8/int4?

[jungyin]

I tried using FP16 and FP32 to store the model separately. The model obtained from FP16 resulted in greater memory consumption (4GB) and slower speed (3token/s) when running with onnxruntime, but using FP32 occupied (3GB) and ran at a faster speed (10token/s). Preliminary assessment suggests that the hardware may not support the FP16 type, resulting in a conversion from FP16 to FP32. How can I determine if my hardware supports types such as FP16, INT8, and INT4

[NarutoUA]

for DirectML you can create dummy device and check DML_FEATURE_DATA_TENSOR_DATA_TYPE_SUPPORT

[github-actions]

This issue has been automatically marked as stale due to inactivity and will be closed in 30 days if no further activity occurs. If further support is needed, please provide an update and/or more details.

[snnn]

This issue has been automatically closed as 'not planned' because it has been marked as 'stale' for more than 30 days without activity. If you believe this is still an issue, please feel free to reopen it.