小行星

softmax后出现nan

Posted on 2019-04-Sun Views:

>>> import torch
>>> import torch.nn.functional as F
>>> F.softmax(torch.Tensor([0, float('-inf')]), -1)
tensor([ 1.0000,  0.0000])
>>> F.softmax(torch.Tensor([0, float('inf')]), -1)  # should give [0.0, 1.0]
tensor([    nan,     nan])
>>> F.log_softmax(torch.Tensor([0, float('-inf')]), -1)
tensor([ 0.0000,    -inf])
>>> F.log_softmax(torch.Tensor([0, float('inf')]), -1)
tensor([    nan,     nan])
>>> F.softmax(torch.Tensor([float('-inf'), 0, float('-inf')]), -1)
tensor([ 0.0000,  1.0000,  0.0000])
>>> F.softmax(torch.Tensor([0, float('inf'), 0]), -1)  # should give [0.0, 1.0, 0.0]
tensor([    nan,     nan,     nan])
>>> F.softmax(torch.Tensor([float('-inf'), 0, float('inf')]), -1)  # should give [0.0, 0.0, 1.0]
tensor([    nan,     nan,     nan])

tf.scan使用

Posted on 2019-04-Sun Views:

设函数为f，
  x = [u(0),u(1),...,u(n)]
  y = tf.scan(f,x,initializer=v(0))
  此时f的参数类型必须是(v(0),x)，f的输出必须和v(0)保持一致，整个计算过程如下：
  v(1)=f(v(0),u(0))
  v(2)=f(v(1),u(1))
  ....
  v(n+1)=f(v(n),u(n))
  y=v(n+1)

python path问题python2和python3

Posted on 2019-04-Sat Views:

python path一直出现问题，
后来把python3降低到python2，问题消失。

python读取帧数

Posted on 2019-04-Thu Views:

import imageio
vid_path = "/data/zhangjunpeng/data/tvsum/video/PJrm840pAUI.mp4"
vid = imageio.get_reader(vid_path, "ffmpeg")
print(vid.count_frames())

重启后 NVIDIA-SMI has failed because it couldn't communicate with the NVIDIA driver

Posted on 2019-03-Sat Views:

这个问题一般是因为linux系统内核小版本号升级导致的…

1 2	sudo cp /etc/default/grub /etc/default/grub.bak sudo vim /etc/default/grub

1 2	GRUB_DEFAULT="Advanced options for Ubuntu>Ubuntu, with Linux 3.13.0-53-generic" #<-修改 sudo update-grub

然后 reboot.

这样操作之后nvidia-smi还是失败状态.

1 2	sudo apt-get install dkms sudo dkms install -m nvidia -v 396-396.54

ubuntu16.04修改grub启动顺序

Posted on 2019-03-Sat Views:

https://askubuntu.com/questions/216398/set-older-kernel-as-default-grub-entry

$ lsb_release -s -rdc
Ubuntu 16.04.4 LTS
16.04
xenial

$ update-grub --version
grub-mkconfig (GRUB) 2.02~beta2-36ubuntu3.17

$ grep -Ei 'submenu|menuentry ' /boot/grub/grub.cfg | sed -re "s/(.? )'([^']+)'.*/\1 \2/"
menuentry  Ubuntu
submenu  Advanced options for Ubuntu
    menuentry  Ubuntu, with Linux 4.4.0-34-generic
    menuentry  Ubuntu, with Linux 4.4.0-34-generic (upstart)
    menuentry  Ubuntu, with Linux 4.4.0-34-generic (recovery mode)
menuentry  System setup

重启后显卡驱动出现问题

Posted on 2019-03-Sat Edited on 2019-06-Tue Views:

1.该命令显示内核的启动顺序

zgw@zgw-ThinkPad:~$ grep menuentry /boot/grub/grub.cfg
if [ x"${feature_menuentry_id}" = xy ]; then
  menuentry_id_option="--id"
  menuentry_id_option=""
export menuentry_id_option
menuentry 'Ubuntu' --class ubuntu --class gnu-linux --class gnu --class os $menuentry_id_option 'gnulinux-simple-5bce3795-da96-4c6f-bed2-67d37185a77d' {
submenu 'Ubuntu 高级选项' $menuentry_id_option 'gnulinux-advanced-5bce3795-da96-4c6f-bed2-67d37185a77d' {
    menuentry 'Ubuntu，Linux 4.8.0-26-lowlatency' --class ubuntu --class gnu-linux --class gnu --class os $menuentry_id_option 'gnulinux-4.4.0-45-lowlatency-advanced-5bce3795-da96-4c6f-bed2-67d37185a77d' {
    menuentry 'Ubuntu, with Linux 4.8.0-26-lowlatency (upstart)' --class ubuntu --class gnu-linux --class gnu --class os $menuentry_id_option 'gnulinux-4.4.0-45-lowlatency-init-upstart-5bce3795-da96-4c6f-bed2-67d37185a77d' {
    menuentry 'Ubuntu, with Linux 4.8.0-26-lowlatency (recovery mode)' --class ubuntu --class gnu-linux --class gnu --class os $menuentry_id_option 'gnulinux-4.4.0-45-lowlatency-recovery-5bce3795-da96-4c6f-bed2-67d37185a77d' {
    menuentry 'Ubuntu，Linux 4.8.0-26-generic' --class ubuntu --class gnu-linux --class gnu --class os $menuentry_id_option 'gnulinux-4.4.0-45-generic-advanced-5bce3795-da96-4c6f-bed2-67d37185a77d' {
    menuentry 'Ubuntu, with Linux 4.8.0-26-generic (upstart)' --class ubuntu --class gnu-linux --class gnu --class os $menuentry_id_option 'gnulinux-4.4.0-45-generic-init-upstart-5bce3795-da96-4c6f-bed2-67d37185a77d' {
    menuentry 'Ubuntu, with Linux 4.8.0-26-generic (recovery mode)' --class ubuntu --class gnu-linux --class gnu --class os $menuentry_id_option 'gnulinux-4.4.0-45-generic-recovery-5bce3795-da96-4c6f-bed2-67d37185a77d' {
    menuentry 'Ubuntu，Linux 4.4.0-21-generic' --class ubuntu --class gnu-linux --class gnu --class os $menuentry_id_option 'gnulinux-4.4.0-21-generic-advanced-5bce3795-da96-4c6f-bed2-67d37185a77d' {
    menuentry 'Ubuntu, with Linux 4.4.0-21-generic (upstart)' --class ubuntu --class gnu-linux --class gnu --class os $menuentry_id_option 'gnulinux-4.4.0-21-generic-init-upstart-5bce3795-da96-4c6f-bed2-67d37185a77d' {
    menuentry 'Ubuntu, with Linux 4.4.0-21-generic (recovery mode)' --class ubuntu --class gnu-linux --class gnu --class os $menuentry_id_option 'gnulinux-4.4.0-21-generic-recovery-5bce3795-da96-4c6f-bed2-67d37185a77d' {
menuentry 'Memory test (memtest86+)' {
menuentry 'Memory test (memtest86+, serial console 115200)' {

假设你要以4.4.0-21内核版本启动，则将文件/etc/default/grub中

　　　　　　Ubuntu, with Linux 4.4.0-72-generic　　

GRUB_DEFAULT=0
改为

GRUB_DEFAULT=6
或者改为

GRUB_DEFAULT=”Ubuntu，Linux 4.4.0-21-generic“
保存后

然后使用命令sudo update-grub

4.重新启动后输入uname -r查看，内核即为想要的内核。

作者：StruggleYouth
来源：CSDN
原文：https://blog.csdn.net/u011304615/article/details/70920171
版权声明：本文为博主原创文章，转载请附上博文链接！

无法连接NVIDIA驱动：NVIDIA-SMI has failed because it couldn't communicate with the NVIDIA driver

Posted on 2019-03-Sat Views:

https://blog.csdn.net/hangzuxi8764/article/details/86572093

model初始化遇到的问题

Posted on 2019-03-Wed Views:

解决:

self.model = Model()
self.model.eval()
self.model.apply(weights_init)

self.block_model = Model()
self.block_model.eval()
self.block_model.apply(weights_init)

训练网络nan总结

Posted on 2019-03-Tue Views:

试着降低lr, 如果还不行就是数据集的问题