QNN toolkit
#874
QNN toolkit
#874
Conversation
|
not checkin the binray files such .raw |
| output_model_path: str, | ||
| ) -> QNNModelHandler: | ||
| main_cmd = "qnn-context-binary-generator" | ||
| runner = QNNSDKRunner(dev=True) |
There was a problem hiding this comment.
what's meaning dev=True for all cases?
There was a problem hiding this comment.
dev=True means run the cmd locally. Run QNN optimization has following steps:
- convert and build model in x64 windows/linux (dev = True
- send the model to target devices(arm-windows or android) (dev=False
There was a problem hiding this comment.
If it is used for run locally or remotely, suggest to change the name to local_run or something else.
In my mind, dev=True means run the optimization using the local branch Olive bits.
There was a problem hiding this comment.
Updated with optional_local_run where the True value is used to tell current runner can run cmd locally.
The value is set to False only if the cmd have to be run remotely but cannot in local env.
olive/model/handler/qnn.py
Outdated
| if not model_paths: | ||
| raise FileNotFoundError(f"No model file found in {model_folder}") | ||
| elif len(model_paths) > 1: | ||
| raise RuntimeError(f"Multiple model files found in {model_folder}: {model_paths}") |
There was a problem hiding this comment.
should the optimization pass since the RuntimeError will be ignored? or need we raise assertionerror so the optimization will fail?
There was a problem hiding this comment.
Make sense. Updated both cases with assertion.
Describe your changes
QNN toolkit
CI e2e test for qnn mobile net: https://github.com/microsoft/Olive/pull/874/checks?check_run_id=20488489220
Checklist before requesting a review
lintrunner -a(Optional) Issue link