Hi, all,
I tested mobilenetv1, v2 and AlexNet on MIX2 (snapdragon835), which CPU, GPU, DSP supported according to device list in SNPE doc. But I found the results of mobilenetv1, v2 are wrong, and AlexNet-fxp-cpu is wrong.
I verify correctness according to the script below `models/alexnet/script/show_alexnet_classification.py`. Any idea? Thanks in advance.
Opinions expressed in the content posted here are the personal opinions of the original authors, and do not necessarily reflect those of Qualcomm Incorporated or its subsidiaries (“Qualcomm”). The content is provided for informational purposes only and is not meant to be an endorsement or representation by Qualcomm or any other party. This site may also provide links or references to non-Qualcomm sites and resources. Qualcomm makes no representations, warranties, or other commitments whatsoever about any non-Qualcomm sites or third-party resources that may be referenced, accessible from, or linked to this site.
Hi ysh329,
It's reported that Mobilenet is vulnerable to quantization. So Qualcomm designed quantization-friendly Mobilenet. Refer to below paper.
https://arxiv.org/pdf/1803.08607.pdf
Thanks,
Jihoonk
Thanks. Do you have paper about current quantization strategy? Thanks in advance
Besides, I found segmentation fault when running mobilenetv1 on Snapdragon835 (MIX 2).
Refer to below document regarding current SNPE quantization.
https://developer.qualcomm.com/docs/snpe/quantized_models.html
Thanks,
jihoonk
hi, not only quanti. but also un-quant are wrong...
have any idea?