-
Notifications
You must be signed in to change notification settings - Fork 3.5k
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
[QNN] Dynamic scale, zero point in qnn.op.dequantize #6849
Conversation
@@ -29,6 +29,9 @@ | |||
import tvm | |||
import tvm.testing | |||
from tvm import relay | |||
from tvm.contrib import graph_runtime |
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.
This is a test script for pytorch frontend... Please move these changes to relay/test_op_qnn_dequantize.py
and relay/test_op_qnn_quantize.py
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.
Moved - thanks for pointing that out
thanks @electriclilies |
* add dynamic dequantize * register quantize and dequantize as opaque * make tests better * black * remove main fn * fix black again * move tests * fix import * fix import again * try again * fix import
* add dynamic dequantize * register quantize and dequantize as opaque * make tests better * black * remove main fn * fix black again * move tests * fix import * fix import again * try again * fix import
* add dynamic dequantize * register quantize and dequantize as opaque * make tests better * black * remove main fn * fix black again * move tests * fix import * fix import again * try again * fix import
This PR allows the qnn.op.dequantize op to accept dynamic scales and zero points. #6782 made similar changes to the qnn.quantize op.
I also added unit tests to the qnn testing to make sure that the dynamic scale and zero points work for both quantize and dequantize.
@jwfromm @masahi Please take a look