search

ArtyZe / yolov3_lite

yolov3 model compress and acceleration (quantization, sparse)， c++ version
37 stars 9 forks source link

AVX Problem #8

Open yukai0406 opened 4 years ago

yukai0406 commented 4 years ago

When I make didn't have the AVX will have error error

ArtyZe commented 4 years ago

@yukai0406 Hello, I have updated batches of updates in my repo, and fixed bugs include your above issue, just have a try :)

yukai0406 commented 4 years ago

I will try

ArtyZe commented 4 years ago

@yukai0406 suggest you pay some attention to quantization, model sparsity will be replaced by NAS in my opinion, and NCNN of Tencent is a better framework if you want to run your model in embedded device instead of darknet

yukai0406 commented 4 years ago

Thank you for the recommendation.

yukai0406 commented 4 years ago

Did you have a plan to add winograd algorithm like ncnn?

ArtyZe commented 4 years ago

@yukai0406 As you seen, I have changed darknet to c++ version and it's convenient to port code of other framwork such as ncnn to darknet. you can have a try. Maybe I will start this work in 5.1 festival if I have free time :)