打开微信,使用扫一扫进入页面后,点击右上角菜单,
点击“发送给朋友”或“分享到朋友圈”完成分享
【寒武纪硬件产品型号】必填*:
必填项,例如:MLU370
【使用操作系统】必填*:
例如:ubuntu
【使用驱动版本】必填*:
例如v4.20.6
【出错信息】 _00000_conv0.conv_in errRate = 0.0
_00000_conv0.conv_out errRate = 0.0
_00001_conv0.bn_in errRate = 0.0
_00001_conv0.bn_out errRate = 0.0
_00002_conv0.act_in errRate = 0.0
_00002_conv0.act_out errRate = 0.0
_00003_conv1.act_in errRate = 0.0
_00003_conv1.act_out errRate = 0.0
_00004_c32.cv1.act_in errRate = 0.0
_00004_c32.cv1.act_out errRate = 0.0
_00005_c32.cv2.act_in errRate = 0.0
_00005_c32.cv2.act_out errRate = 0.0
_00006_c32.cv3.act_in errRate = 0.0
_00006_c32.cv3.act_out errRate = 0.0
_00007_c32.m.0.cv1.act_in errRate = 0.0
_00007_c32.m.0.cv1.act_out errRate = 0.0
_00008_c32.m.0.cv2.act_in errRate = 0.0
_00008_c32.m.0.cv2.act_out errRate = 0.0
_00009_conv3.act_in errRate = 0.0
_00009_conv3.act_out errRate = 0.0
_00010_c34.cv1.act_in errRate = 0.0
_00010_c34.cv1.act_out errRate = 0.0
_00011_c34.cv2.act_in errRate = 0.0
_00011_c34.cv2.act_out errRate = 0.0
_00012_c34.cv3.act_in errRate = 0.0
_00012_c34.cv3.act_out errRate = 0.0
_00013_c34.m.0.cv1.act_in errRate = 0.0
_00013_c34.m.0.cv1.act_out errRate = 0.0
_00014_c34.m.0.cv2.act_in errRate = 0.0
_00014_c34.m.0.cv2.act_out errRate = 0.0
_00015_c34.m.1.cv1.act_in errRate = 0.0
_00015_c34.m.1.cv1.act_out errRate = 0.0
_00016_c34.m.1.cv2.act_in errRate = 0.0
_00016_c34.m.1.cv2.act_out errRate = 0.0
。。。。。。。。。。。。 中间的省略了有最大字数限制
_00124_conv0.act_in errRate = 0.0
_00124_conv0.act_out errRate = 0.0
_00125_conv1.act_in errRate = 0.0
_00125_conv1.act_out errRate = 0.0
_00126_c32.cv1.act_in errRate = 0.0
_00126_c32.cv1.act_out errRate = 0.0
_00127_c32.cv2.act_in errRate = 0.0
_00127_c32.cv2.act_out errRate = 0.0
_00128_c32.cv3.act_in errRate = 0.0
_00128_c32.cv3.act_out errRate = 0.0
_00129_c32.m.0.cv1.act_in errRate = 0.0
_00129_c32.m.0.cv1.act_out errRate = 0.0
_00130_c32.m.0.cv2.act_in errRate = 0.0
_00130_c32.m.0.cv2.act_out errRate = 0.0
_00131_conv3.act_in errRate = 0.0
_00131_conv3.act_out errRate = 0.0
_00132_c34.cv1.act_in errRate = 0.0
_00132_c34.cv1.act_out errRate = 0.0
_00133_c34.cv2.act_in errRate = 0.0
_00133_c34.cv2.act_out errRate = 0.0
_00134_c34.cv3.act_in errRate = 0.0
_00134_c34.cv3.act_out errRate = 0.0
_00135_c34.m.0.cv1.act_in errRate = 0.0
_00135_c34.m.0.cv1.act_out errRate = 0.0
_00136_c34.m.0.cv2.act_in errRate = 0.0
.。。。。。。。
1、为什么会重复出现?出现一次conv0 之后还会在出现 多次conv0
2、这个地方显示的误差是0,但是我实际在测的时候,cpu推理和mlu逐层推理之间存在很大的误差?
这个到底该怎么使用?
热门帖子
精华帖子