1)ctrl+Z:停止當前進程
首先先將一個程序運行起來,這個時候如果你需要去干別的事情,需要暫停運行,可以使用ctrl+Z:
user@mine:/opt/user/pytorch-gender$ python train_debug.py --debugFile=./debug {'debugFile': './debug'} Epoch 0/199 ---------- ^Z [1]+ Stopped python train_debug.py --debugFile=./debug
從上面可以看見,這個程序已經已經停止了,狀態為Stopped
2)jobs:用於查看正在運行的命令
user@mine:/opt/user/pytorch-gender$ jobs [1]+ Stopped python train_debug.py --debugFile=./debug
前面的編號[1]是命令編號
3)bg 命令編號:把程序調度到后台執行:
user@mine:/opt/user/pytorch-gender$ bg 1 [1]+ python train_debug.py --debugFile=./debug & user@mine:/opt/user/pytorch-gender$ jobs [1]+ Running python train_debug.py --debugFile=./debug &
然后我們可以看見該命令在后台運行起來了,狀態為Running,命令后的&標志就是把命令放在后台運行的意思
這個時候該命令生成的返回信息會自己打印出來:
user@mine:/opt/user/pytorch-gender$ train Loss: 0.4611 Acc: 0.7824 val Loss: 0.1882 Acc: 0.9340 Epoch 1/199 ---------- user@mine:/opt/user/pytorch-gender$ train Loss: 0.3271 Acc: 0.8578 val Loss: 0.1845 Acc: 0.9260 Epoch 2/199 ----------
不影響你運行其他的命令,你就輸入你的命令回車即可
當然,如果你不想讓輸出顯示在控制台中,那就在運行時指明將輸出信息寫入日志文件:
user@mine:/opt/user/pytorch-gender$ python train_debug.py --debugFile=./debug >> gender_log_debug_1.out
打開另一個窗口查看日志文件為:
user@mine:/opt/user/pytorch-gender$ cat gender_log_debug_1.out {'debugFile': './debug'} Epoch 0/199 ----------
然后這個時候如果你想進行調試,即pytorch Debug —交互式調試工具Pdb (ipdb是增強版的pdb)-1-在pytorch中使用,那么你在本地生成文件夾debug后,再查看日志文件變為:
user@mine:/opt/user/pytorch-gender$ cat gender_log_debug_1.out {'debugFile': './debug'} Epoch 0/199 ---------- train Loss: 0.4507 Acc: 0.7919 val Loss: 0.1578 Acc: 0.9420 Epoch 1/199 ---------- train Loss: 0.3201 Acc: 0.8576 val Loss: 0.1069 Acc: 0.9540 Epoch 2/199 ---------- --Call-- > /home/mine/anaconda3/lib/python3.6/site-packages/torch/autograd/grad_mode.py(129)__exit__() 128 --> 129 def __exit__(self, *args): 130 torch.set_grad_enabled(self.prev) ipdb> user@mine:/opt/user/pytorch-gender$
這時候你在命令端輸入調試命令u:
user@mine:/opt/user/pytorch-gender$ python train_debug.py --debugFile=./debug >> gender_log_debug_1.out u
可見日志文件中變為:
user@mine:/opt/user/pytorch-gender$ cat gender_log_debug_1.out {'debugFile': './debug'} Epoch 0/199 ---------- train Loss: 0.4507 Acc: 0.7919 val Loss: 0.1578 Acc: 0.9420 Epoch 1/199 ---------- train Loss: 0.3201 Acc: 0.8576 val Loss: 0.1069 Acc: 0.9540 Epoch 2/199 ---------- --Call-- > /home/mine/anaconda3/lib/python3.6/site-packages/torch/autograd/grad_mode.py(129)__exit__() 128 --> 129 def __exit__(self, *args): 130 torch.set_grad_enabled(self.prev) ipdb> > /opt/user/pytorch-gender/train_debug.py(151)train_model() 150 import ipdb; --> 151 ipdb.set_trace() 152 ipdb>
如果調用l 123命令:
user@mine:/opt/user/pytorch-gender$ python train_debug.py --debugFile=./debug >> gender_log_debug_1.out u l 123
可見日志文件又變為:
user@mine:/opt/user/pytorch-gender$ cat gender_log_debug_1.out {'debugFile': './debug'} Epoch 0/199 ---------- train Loss: 0.4507 Acc: 0.7919 val Loss: 0.1578 Acc: 0.9420 Epoch 1/199 ---------- train Loss: 0.3201 Acc: 0.8576 val Loss: 0.1069 Acc: 0.9540 Epoch 2/199 ---------- --Call-- > /home/mine/anaconda3/lib/python3.6/site-packages/torch/autograd/grad_mode.py(129)__exit__() 128 --> 129 def __exit__(self, *args): 130 torch.set_grad_enabled(self.prev) ipdb> > /opt/user/pytorch-gender/train_debug.py(151)train_model() 150 import ipdb; --> 151 ipdb.set_trace() 152 ipdb> 118 labels = labels.to(device) # 當前批次的標簽輸入 119 # print('input : ', inputs) 120 # print('labels : ', labels) 121 122 # 將梯度參數歸0 123 optimizer.zero_grad() 124 125 # 前向計算 126 # track history if only in train 127 with torch.set_grad_enabled(phase == 'train'): 128 # 相應輸入對應的輸出 ipdb>
所以輸出和命令輸入雖然不在一起,但是並不妨礙功能的實現
4)fg 命令編號:將后台命令調到前台運行
如果我想要對上面的命令進行調試,我就需要將其調到前台,然后再進行調試
user@mine:/opt/user/pytorch-gender$ fg 1 python train_debug.py --debugFile=./debug train Loss: 0.2337 Acc: 0.9059 val Loss: 0.1347 Acc: 0.9400 Epoch 4/199 ---------- train Loss: 0.2040 Acc: 0.9141 val Loss: 0.0962 Acc: 0.9640 Epoch 5/199 ---------- train Loss: 0.1984 Acc: 0.9182 val Loss: 0.0825 Acc: 0.9720 Epoch 6/199 ---------- train Loss: 0.1841 Acc: 0.9218 val Loss: 0.1059 Acc: 0.9640 Epoch 7/199 ---------- train Loss: 0.1868 Acc: 0.9215 val Loss: 0.0668 Acc: 0.9740 Epoch 8/199 ---------- train Loss: 0.1782 Acc: 0.9273 val Loss: 0.0735 Acc: 0.9740 Epoch 9/199 ---------- train Loss: 0.1703 Acc: 0.9291 val Loss: 0.0850 Acc: 0.9680 Epoch 10/199 ---------- train Loss: 0.1596 Acc: 0.9329 val Loss: 0.1114 Acc: 0.9560 Epoch 11/199 ---------- --Call-- > /home/mine/anaconda3/lib/python3.6/site-packages/torch/autograd/grad_mode.py(129)__exit__() 128 --> 129 def __exit__(self, *args): 130 torch.set_grad_enabled(self.prev) ipdb>
5)nohup 命令 & :直接將命令放在后台運行
nohup python train_debug.py --debugFile=./debug &
如果要指定返回信息寫入的日志文件log.out:
nohup python train_debug.py --debugFile=./debug >> log.out &