BUG1989/caffe-int8-convert-tools

Issues

• 请问在哪些硬件平台可以验证有加速？

  #68 opened 3 years ago by JensenHJS
  0

• Can ncnn2caffe and how to.

  #60 opened 5 years ago by zhiwei-dong
  1

• 能提供下例子中的caffe模型吗

  #67 opened 4 years ago by ZhiqiJiang
  0

• 请问这个不能量化mobilenetyolov3吗

  #53 opened 5 years ago by MaxwellAlen
  5

• 得到table文件后，想转成int8的caffemodel文件

  #51 opened 5 years ago by shonehang
  11

• A question about "Test Dataset"

  #66 opened 4 years ago by bit2erswe2t
  0

• threshold_distribution中的代码问题

  #43 opened 5 years ago by tianylijun
  5

• 支持yolo量化吗？

  #47 opened 5 years ago by shuxiao9058
  2

• mxnet quantization code not found?

  #65 opened 4 years ago by infrontofme
  0

• 首尾层做量化吗？

  #64 opened 5 years ago by iamhankai
  1

• 使用最新脚本模型加载仍然报错

  #63 opened 5 years ago by grainw
  0

• NameError: global name 'layer' is not defined

  #62 opened 5 years ago by NJUSTghw
  0

• 请问作者试过量化语义分割网络模型吗？

  #61 opened 5 years ago by duzeyan
  1

• pytorch 版本报错 'dict' object has no attribute 'named_modules'

  #58 opened 5 years ago by Enshall
  2

• Does the table file support the integer quantized Tensor-flow model in the future?

  #59 opened 5 years ago by dainHandle
  0

• How about using float16 with neon ?

  #57 opened 5 years ago by peyer
  0

• INT8 量化后的分类网络，不同图片，分类结果一样

  #56 opened 5 years ago by SophieChang66
  0

• 模型加载报错

  #55 opened 5 years ago by jianbaba
  1

• 请问faster-rcnn的VGG16模型怎么使用这个工具压缩？

  #50 opened 5 years ago by baizhenmao95
  2

• TypeError: unorderable types: int() > str()

  #45 opened 5 years ago by 1343464520
  3

• 为何量化完 都为inf

  #52 opened 5 years ago by popper0912
  1

• 报个case：量化后某个层的weight scale全为0

  #42 opened 5 years ago by dancingpipi
  1

• QUANTIZE_WINOGRAND_NUM 的作用

  #41 opened 5 years ago by sxsxsx
  1

• mtcnn量化后精度下降严重

  #49 opened 5 years ago by ivoxx
  0

• 怎么在caffe上测试准确率呢？

  #46 opened 5 years ago by andeyeluguo
  0

• readme里面的implementation是不是写错了

  #48 opened 5 years ago by andeyeluguo
  0

• Converted Mobilefacenet_Res model, low accuracy

  #36 opened 6 years ago by WIll-Xu35
  17

• 量化时对batchnorm层以及包含bias的conv层的处理方法？

  #44 opened 5 years ago by CrawlingD
  1

• network_prepare

  #40 opened 5 years ago by shuyanzhu
  0

• Does the version of protobuf affect the results?

  #39 opened 5 years ago by BladeRunner2050
  2

• 量化后变得更慢了？

  #34 opened 6 years ago by sonixixi
  4

• 自己使用此工具对Squeezenet模型进行int8转换遇到问题

  #20 opened 5 years ago by JSnobody
  16

• tensorflow2ncnn 对pb有量化操作么

  #37 opened 6 years ago by dachengai
  0

• arm v7 上 量化模型中 卷积层 消耗的时间，有个规律 卷积核 个数越多的卷积层，量化模型这一卷积层时间消耗降低越明显。卷积核个数越少的话，量化模型这一卷积层时间消耗相比float32模型，基本没有变化，甚至时间消耗增加。

  #35 opened 6 years ago by sonixixi
  1

• Conv time of squeezenet-int8

  #33 opened 6 years ago by SecretGuardian
  1

• 只对卷积核为3x3和1x1的作了量化？

  #29 opened 6 years ago by xiangyu19
  2

• raise ValueError("First layer should be input")

  #30 opened 6 years ago by suikammd
  1

• Why the range of blob_distubution is set to (0,th) in caffe-int8-convert-tool-dev.py:116

  #32 opened 6 years ago by bin913
  1

• 运行后有错误。

  #31 opened 6 years ago by Andrawzyf
  1

• 校准权重矩阵

  #28 opened 6 years ago by shenghsiaowong
  1

• KL计算时疑问

  #26 opened 6 years ago by zw1992
  3

• 关于vgg-ssd转化的几个问题

  #21 opened 6 years ago by zhangyunming
  11

• 网络压缩

  #24 opened 6 years ago by shenghsiaowong
  3

• How to quantize BatchNorm？

  #27 opened 6 years ago by jiangwx
  4

• Questions about the accuracy

  #25 opened 6 years ago by jycccccccc
  0

• mobilenet_v1_ssd_weight.table

  #23 opened 6 years ago by shanshan0412
  1

• pc测试，int8速度降低一半

  #22 opened 6 years ago by lvpiche
  4

• do this tools upport darknet int8 convert

  #17 opened 6 years ago by azuryl
  1

• 关于此工具量化与TensorRT量化

  #18 opened 6 years ago by JSnobody
  1

• Caffe or NCNN？

  #19 opened 6 years ago by JSnobody
  2