-
Notifications
You must be signed in to change notification settings - Fork 8
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Added QNN EP for ARM64 builds. #41
base: main
Are you sure you want to change the base?
Conversation
Not ready for review since the exported projects are not great yet... I'm not sure what to do there... |
@@ -23,7 +23,8 @@ public static ModelCompatibility GetModelCompatibility(ModelDetails modelDetails | |||
|
|||
ModelCompatibilityState compatibility; | |||
if (modelDetails.HardwareAccelerators.Contains(HardwareAccelerator.CPU) || | |||
(modelDetails.HardwareAccelerators.Contains(HardwareAccelerator.DML) && modelDetails.SupportedOnQualcomm == true)) | |||
(modelDetails.HardwareAccelerators.Contains(HardwareAccelerator.DML) && modelDetails.SupportedOnQualcomm == true) || |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
We should disable all DML models on Arm64
{ | ||
Dictionary<string, string> options = new() | ||
{ | ||
{ "backend_path", "QnnHtp.dll" } |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
we should include the high performance kv here too
@@ -18,7 +18,8 @@ | |||
"Description": "This is an embedding model executed on the CPU or GPU", | |||
"HardwareAccelerator": [ | |||
"CPU", | |||
"DML" | |||
"DML", | |||
"QNN" |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
we need to do some testing here before enabling the models for qnn - I'm seeing both the cpu and npu running which might mean the model is falling back to cpu and it's not optimized for qnn
No description provided.