-
您好,我在利用benchncnn测试模型遇到一些速度问题
测试平台: linux服务器、骁龙845芯片
ncn版本: 20200727
编译链接:https://github.com/Tencent/ncnn/wiki/how-to-build#build-for-linux-x86
https://github.com/Tencent/ncnn/wiki/how-to-build#b…
qinb updated
3 months ago
-
Hey, simply running:
`python main.py --mode prune --model nasnet --batch-size 128 --restore True --dataset cifar100 --method random --speed-up 10 --global-pruning`
Doesn't work to prune Nasnet. I…
-
实测MobileNetV2的FLOPs只有MobileNetV1的一半,ShuffleNetV2更少一些,但是无论是so还是tensorRT,ShuffleNet和MobileNet的infer时间都会更长。另外,最近测试mnasnet和EfficientNet,速度也不行,在FLOPs明显降低很多的情况下速度还是太慢。还有一点,在非relu激活函数的网络里,tensorRT往往表现的比so预测更…
-
## detail | 详细描述 | 詳細な説明
loop_count = 64
num_threads = 8
powersave = 0
gpu_device = -1
cooling_down = 0
squeezenet min = 20.73 max = 142.01 avg = 47.55
squeezenet_int8 m…
-
## 📚 Documentation
Hi!
The following models are listed in our [hubconf.py](https://github.com/pytorch/vision/blob/master/hubconf.py)
but these models are not listed over [pytorch hub](https://p…
-
# 모델별 학습 기록(Acc/F1)
- https://nayohan.notion.site/07-30bd7e17f7f44b078d22342d21131d65
## 1. Torchvision models
- [ ] VGG-19 with batch normalization
- [x] `ResNet-18`: class weight, bactch siz…
-
## Why
Deep Learning 輪講は最新の技術や論文を追うことで、エンジニアが「技術で解決できること」のレベルをあげていくことを目的にした会です。
prev. https://github.com/wantedly/machine-learning-round-table/issues/1
## What
話したいことがある人はここにコメントしましょう!
面白…
-
I copied the code below from the Mixnet model builder code, trying to use it as a "hello world"
before adapting the builder to my own problem. This code should build a mixnet_s, IIUC.
I am running i…
-
This is a master issue to track requests for adding new pre-trained models to torchvision.
Here is the (potentially incomplete) list I compiled:
- [x] ResNext https://github.com/pytorch/vision/iss…
-
- mix2,高通835
## 高通835-8线程-armv7
```shell
chiron:/data/local/tmp/ncnn $ ./benchncnn …