In [2]:
# memory footprint support libraries/code
!ln -sf /opt/bin/nvidia-smi /usr/bin/nvidia-smi
!pip install gputil
!pip install psutil
!pip install humanize
import psutil
import humanize
import os
import GPUtil as GPU
GPUs = GPU.getGPUs()
# XXX: only one GPU on Colab and isn’t guaranteed
gpu = GPUs[0]
def printm():
process = psutil.Process(os.getpid())
print("Gen RAM Free: " + humanize.naturalsize( psutil.virtual_memory().available ), " | Proc size: " + humanize.naturalsize( process.memory_info().rss))
print("GPU RAM Free: {0:.0f}MB | Used: {1:.0f}MB | Util {2:3.0f}% | Total {3:.0f}MB".format(gpu.memoryFree, gpu.memoryUsed, gpu.memoryUtil*100, gpu.memoryTotal))
printm()
Requirement already satisfied: gputil in /usr/local/lib/python2.7/dist-packages (1.3.0)
Requirement already satisfied: numpy in /usr/local/lib/python2.7/dist-packages (from gputil) (1.14.5)
Requirement already satisfied: psutil in /usr/local/lib/python2.7/dist-packages (5.4.6)
Requirement already satisfied: humanize in /usr/local/lib/python2.7/dist-packages (0.5.1)
('Gen RAM Free: 12.6 GB', ' | Proc size: 150.3 MB')
GPU RAM Free: 11438MB | Used: 1MB | Util 0% | Total 11439MB
In [4]:
!apt install libnvrtc8.0
!pip install mxnet-cu80
import mxnet as mx
Reading package lists... Done
Building dependency tree
Reading state information... Done
The following NEW packages will be installed:
libnvrtc8.0
0 upgraded, 1 newly installed, 0 to remove and 0 not upgraded.
Need to get 6,225 kB of archives.
After this operation, 28.3 MB of additional disk space will be used.
Get:1 http://archive.ubuntu.com/ubuntu artful/multiverse amd64 libnvrtc8.0 amd64 8.0.61-1 [6,225 kB]
Fetched 6,225 kB in 1s (5,443 kB/s)
78Selecting previously unselected package libnvrtc8.0:amd64.
(Reading database ... 18396 files and directories currently installed.)
Preparing to unpack .../libnvrtc8.0_8.0.61-1_amd64.deb ...
7Progress: [ 0%] [..........................................................] 87Progress: [ 16%] [#########.................................................] 8Unpacking libnvrtc8.0:amd64 (8.0.61-1) ...
7Progress: [ 33%] [###################.......................................] 87Progress: [ 50%] [#############################.............................] 8Setting up libnvrtc8.0:amd64 (8.0.61-1) ...
7Progress: [ 66%] [######################################....................] 87Progress: [ 83%] [################################################..........] 8Processing triggers for libc-bin (2.26-0ubuntu2.1) ...
78Collecting mxnet-cu80
Downloading https://files.pythonhosted.org/packages/f6/6c/566a1d4b8b1005b7d9ccfaecd7632f6dca596246f6657827b2d4e97c72c7/mxnet_cu80-1.2.1-py2.py3-none-manylinux1_x86_64.whl (299.1MB)
23% |███████▌ | 70.3MB 27.8MB/s eta 0:00:09 100% |████████████████████████████████| 299.1MB 58kB/s
Collecting graphviz<0.9.0,>=0.8.1 (from mxnet-cu80)
Downloading https://files.pythonhosted.org/packages/53/39/4ab213673844e0c004bed8a0781a0721a3f6bb23eb8854ee75c236428892/graphviz-0.8.4-py2.py3-none-any.whl
Requirement already satisfied: numpy<1.15.0,>=1.8.2 in /usr/local/lib/python2.7/dist-packages (from mxnet-cu80) (1.14.5)
Requirement already satisfied: requests<2.19.0,>=2.18.4 in /usr/local/lib/python2.7/dist-packages (from mxnet-cu80) (2.18.4)
Requirement already satisfied: idna<2.7,>=2.5 in /usr/local/lib/python2.7/dist-packages (from requests<2.19.0,>=2.18.4->mxnet-cu80) (2.6)
Requirement already satisfied: urllib3<1.23,>=1.21.1 in /usr/local/lib/python2.7/dist-packages (from requests<2.19.0,>=2.18.4->mxnet-cu80) (1.22)
Requirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python2.7/dist-packages (from requests<2.19.0,>=2.18.4->mxnet-cu80) (2018.4.16)
Requirement already satisfied: chardet<3.1.0,>=3.0.2 in /usr/local/lib/python2.7/dist-packages (from requests<2.19.0,>=2.18.4->mxnet-cu80) (3.0.4)
Installing collected packages: graphviz, mxnet-cu80
Successfully installed graphviz-0.8.4 mxnet-cu80-1.2.1
In [8]:
!git clone https://github.com/hzy46/Char-RNN-TensorFlow.git
Cloning into 'Char-RNN-TensorFlow'...
remote: Counting objects: 26, done.
remote: Total 26 (delta 0), reused 0 (delta 0), pack-reused 26
Unpacking objects: 100% (26/26), done.
In [9]:
!cd Char-RNN-TensorFlow && python train.py --use_embedding --input_file data/poetry.txt --name poetry --learning_rate 0.005 --num_steps 26 --num_seqs 32 --max_steps 10000
5387
3501
WARNING:tensorflow:From /content/Char-RNN-TensorFlow/model.py:93: softmax_cross_entropy_with_logits (from tensorflow.python.ops.nn_ops) is deprecated and will be removed in a future version.
Instructions for updating:
Future major versions of TensorFlow will allow gradients to flow
into the labels input on backprop by default.
See @{tf.nn.softmax_cross_entropy_with_logits_v2}.
2018-07-28 14:28:23.339235: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:897] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2018-07-28 14:28:23.339827: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1392] Found device 0 with properties:
name: Tesla K80 major: 3 minor: 7 memoryClockRate(GHz): 0.8235
pciBusID: 0000:00:04.0
totalMemory: 11.17GiB freeMemory: 11.10GiB
2018-07-28 14:28:23.339881: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1471] Adding visible gpu devices: 0
2018-07-28 14:28:23.800867: I tensorflow/core/common_runtime/gpu/gpu_device.cc:952] Device interconnect StreamExecutor with strength 1 edge matrix:
2018-07-28 14:28:23.800958: I tensorflow/core/common_runtime/gpu/gpu_device.cc:958] 0
2018-07-28 14:28:23.800987: I tensorflow/core/common_runtime/gpu/gpu_device.cc:971] 0: N
2018-07-28 14:28:23.801406: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1084] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 10763 MB memory) -> physical GPU (device: 0, name: Tesla K80, pci bus id: 0000:00:04.0, compute capability: 3.7)
step: 10/10000... loss: 6.5503... 0.0976 sec/batch
step: 20/10000... loss: 6.5084... 0.0995 sec/batch
step: 30/10000... loss: 6.3076... 0.0980 sec/batch
step: 40/10000... loss: 6.2250... 0.1381 sec/batch
step: 50/10000... loss: 6.0408... 0.0891 sec/batch
step: 60/10000... loss: 5.8749... 0.0840 sec/batch
step: 70/10000... loss: 5.7850... 0.0878 sec/batch
step: 80/10000... loss: 5.7318... 0.0983 sec/batch
step: 90/10000... loss: 5.6727... 0.0870 sec/batch
step: 100/10000... loss: 5.5724... 0.0877 sec/batch
step: 110/10000... loss: 5.5872... 0.0835 sec/batch
step: 120/10000... loss: 5.4830... 0.0966 sec/batch
step: 130/10000... loss: 5.4636... 0.0872 sec/batch
step: 140/10000... loss: 5.4168... 0.0886 sec/batch
step: 150/10000... loss: 5.4805... 0.1149 sec/batch
step: 160/10000... loss: 5.4889... 0.0846 sec/batch
step: 170/10000... loss: 5.4045... 0.0937 sec/batch
step: 180/10000... loss: 5.4422... 0.0905 sec/batch
step: 190/10000... loss: 5.4532... 0.1111 sec/batch
step: 200/10000... loss: 5.3409... 0.0842 sec/batch
step: 210/10000... loss: 5.3405... 0.0884 sec/batch
step: 220/10000... loss: 5.4533... 0.0874 sec/batch
step: 230/10000... loss: 5.3560... 0.0832 sec/batch
step: 240/10000... loss: 5.3247... 0.0952 sec/batch
step: 250/10000... loss: 5.3851... 0.0936 sec/batch
step: 260/10000... loss: 5.4183... 0.0867 sec/batch
step: 270/10000... loss: 5.2965... 0.0878 sec/batch
step: 280/10000... loss: 5.3577... 0.0836 sec/batch
step: 290/10000... loss: 5.3775... 0.0930 sec/batch
step: 300/10000... loss: 5.3400... 0.1183 sec/batch
step: 310/10000... loss: 5.3416... 0.0849 sec/batch
step: 320/10000... loss: 5.3459... 0.0811 sec/batch
step: 330/10000... loss: 5.3807... 0.1006 sec/batch
step: 340/10000... loss: 5.4062... 0.0838 sec/batch
step: 350/10000... loss: 5.3606... 0.0880 sec/batch
step: 360/10000... loss: 5.3277... 0.0944 sec/batch
step: 370/10000... loss: 5.3535... 0.0907 sec/batch
step: 380/10000... loss: 5.2762... 0.0889 sec/batch
step: 390/10000... loss: 5.4210... 0.0928 sec/batch
step: 400/10000... loss: 5.3872... 0.0838 sec/batch
step: 410/10000... loss: 5.3875... 0.1104 sec/batch
step: 420/10000... loss: 5.3337... 0.0878 sec/batch
step: 430/10000... loss: 5.3171... 0.0893 sec/batch
step: 440/10000... loss: 5.1649... 0.0990 sec/batch
step: 450/10000... loss: 5.3437... 0.1052 sec/batch
step: 460/10000... loss: 5.3547... 0.0838 sec/batch
step: 470/10000... loss: 5.2432... 0.0885 sec/batch
step: 480/10000... loss: 5.3419... 0.0849 sec/batch
step: 490/10000... loss: 5.2243... 0.0871 sec/batch
step: 500/10000... loss: 5.2685... 0.0850 sec/batch
step: 510/10000... loss: 5.3271... 0.0918 sec/batch
step: 520/10000... loss: 5.3210... 0.0888 sec/batch
step: 530/10000... loss: 5.2131... 0.0957 sec/batch
step: 540/10000... loss: 5.2771... 0.1139 sec/batch
step: 550/10000... loss: 5.3214... 0.0960 sec/batch
step: 560/10000... loss: 5.2217... 0.1049 sec/batch
step: 570/10000... loss: 5.3000... 0.0862 sec/batch
step: 580/10000... loss: 5.2852... 0.0858 sec/batch
step: 590/10000... loss: 5.2651... 0.0832 sec/batch
step: 600/10000... loss: 5.3222... 0.1118 sec/batch
step: 610/10000... loss: 5.1769... 0.0898 sec/batch
step: 620/10000... loss: 5.1711... 0.0859 sec/batch
step: 630/10000... loss: 5.1479... 0.0899 sec/batch
step: 640/10000... loss: 5.2436... 0.0904 sec/batch
step: 650/10000... loss: 5.1731... 0.0952 sec/batch
step: 660/10000... loss: 5.2951... 0.0946 sec/batch
step: 670/10000... loss: 5.3889... 0.0833 sec/batch
step: 680/10000... loss: 5.2643... 0.0886 sec/batch
step: 690/10000... loss: 5.3085... 0.0862 sec/batch
step: 700/10000... loss: 5.2041... 0.0856 sec/batch
step: 710/10000... loss: 5.1424... 0.0882 sec/batch
step: 720/10000... loss: 5.2828... 0.0935 sec/batch
step: 730/10000... loss: 5.1889... 0.0878 sec/batch
step: 740/10000... loss: 5.1116... 0.0879 sec/batch
step: 750/10000... loss: 5.1667... 0.0860 sec/batch
step: 760/10000... loss: 5.1459... 0.0914 sec/batch
step: 770/10000... loss: 5.1754... 0.0906 sec/batch
step: 780/10000... loss: 5.2236... 0.0918 sec/batch
step: 790/10000... loss: 5.1612... 0.1019 sec/batch
step: 800/10000... loss: 5.2516... 0.0903 sec/batch
step: 810/10000... loss: 5.0950... 0.0884 sec/batch
step: 820/10000... loss: 5.2079... 0.0950 sec/batch
step: 830/10000... loss: 5.1323... 0.0876 sec/batch
step: 840/10000... loss: 5.1534... 0.0880 sec/batch
step: 850/10000... loss: 5.0890... 0.0973 sec/batch
step: 860/10000... loss: 5.2005... 0.0959 sec/batch
step: 870/10000... loss: 5.1709... 0.0931 sec/batch
step: 880/10000... loss: 5.2006... 0.0898 sec/batch
step: 890/10000... loss: 5.1437... 0.0962 sec/batch
step: 900/10000... loss: 5.1160... 0.0852 sec/batch
step: 910/10000... loss: 5.0154... 0.0945 sec/batch
step: 920/10000... loss: 5.0621... 0.1522 sec/batch
step: 930/10000... loss: 5.1586... 0.1001 sec/batch
step: 940/10000... loss: 5.2192... 0.0860 sec/batch
step: 950/10000... loss: 5.1639... 0.0880 sec/batch
step: 960/10000... loss: 5.0838... 0.0962 sec/batch
step: 970/10000... loss: 5.0600... 0.0949 sec/batch
step: 980/10000... loss: 5.1152... 0.0951 sec/batch
step: 990/10000... loss: 5.0260... 0.1078 sec/batch
step: 1000/10000... loss: 5.0995... 0.0903 sec/batch
step: 1010/10000... loss: 5.0713... 0.0963 sec/batch
step: 1020/10000... loss: 5.0541... 0.1032 sec/batch
step: 1030/10000... loss: 5.1176... 0.0915 sec/batch
step: 1040/10000... loss: 5.0614... 0.0972 sec/batch
step: 1050/10000... loss: 4.9866... 0.0953 sec/batch
step: 1060/10000... loss: 5.0217... 0.0946 sec/batch
step: 1070/10000... loss: 4.9625... 0.0909 sec/batch
step: 1080/10000... loss: 5.0197... 0.0886 sec/batch
step: 1090/10000... loss: 5.0521... 0.0899 sec/batch
step: 1100/10000... loss: 5.0207... 0.0934 sec/batch
step: 1110/10000... loss: 5.0009... 0.0979 sec/batch
step: 1120/10000... loss: 5.1143... 0.0881 sec/batch
step: 1130/10000... loss: 5.0718... 0.1047 sec/batch
step: 1140/10000... loss: 5.0742... 0.0949 sec/batch
step: 1150/10000... loss: 5.0110... 0.1082 sec/batch
step: 1160/10000... loss: 5.0153... 0.0920 sec/batch
step: 1170/10000... loss: 5.1940... 0.0940 sec/batch
step: 1180/10000... loss: 5.0073... 0.0854 sec/batch
step: 1190/10000... loss: 5.1320... 0.1123 sec/batch
step: 1200/10000... loss: 5.1609... 0.0894 sec/batch
step: 1210/10000... loss: 4.9907... 0.0939 sec/batch
step: 1220/10000... loss: 4.9070... 0.0868 sec/batch
step: 1230/10000... loss: 5.2041... 0.1090 sec/batch
step: 1240/10000... loss: 5.1402... 0.0878 sec/batch
step: 1250/10000... loss: 5.0746... 0.0865 sec/batch
step: 1260/10000... loss: 5.0652... 0.0902 sec/batch
step: 1270/10000... loss: 5.0868... 0.1120 sec/batch
step: 1280/10000... loss: 5.0157... 0.0906 sec/batch
step: 1290/10000... loss: 5.1207... 0.1009 sec/batch
step: 1300/10000... loss: 5.0094... 0.0916 sec/batch
step: 1310/10000... loss: 4.9853... 0.1126 sec/batch
step: 1320/10000... loss: 5.1029... 0.0966 sec/batch
step: 1330/10000... loss: 5.1057... 0.0934 sec/batch
step: 1340/10000... loss: 4.9870... 0.0854 sec/batch
step: 1350/10000... loss: 4.9766... 0.1114 sec/batch
step: 1360/10000... loss: 4.9696... 0.0882 sec/batch
step: 1370/10000... loss: 4.8621... 0.0877 sec/batch
step: 1380/10000... loss: 5.0154... 0.1017 sec/batch
step: 1390/10000... loss: 4.8453... 0.1022 sec/batch
step: 1400/10000... loss: 4.8987... 0.0897 sec/batch
step: 1410/10000... loss: 5.0550... 0.0949 sec/batch
step: 1420/10000... loss: 5.0368... 0.0933 sec/batch
step: 1430/10000... loss: 5.0870... 0.0867 sec/batch
step: 1440/10000... loss: 4.9731... 0.0971 sec/batch
step: 1450/10000... loss: 4.9920... 0.0946 sec/batch
step: 1460/10000... loss: 5.0103... 0.0936 sec/batch
step: 1470/10000... loss: 4.9787... 0.1224 sec/batch
step: 1480/10000... loss: 4.9669... 0.0875 sec/batch
step: 1490/10000... loss: 4.9396... 0.0939 sec/batch
step: 1500/10000... loss: 4.8835... 0.0867 sec/batch
step: 1510/10000... loss: 5.0471... 0.1005 sec/batch
step: 1520/10000... loss: 4.9288... 0.0904 sec/batch
step: 1530/10000... loss: 4.9994... 0.0886 sec/batch
step: 1540/10000... loss: 4.9793... 0.0917 sec/batch
step: 1550/10000... loss: 4.9663... 0.1047 sec/batch
step: 1560/10000... loss: 4.9698... 0.0976 sec/batch
step: 1570/10000... loss: 4.9367... 0.0986 sec/batch
step: 1580/10000... loss: 4.9327... 0.0937 sec/batch
step: 1590/10000... loss: 4.8923... 0.0935 sec/batch
step: 1600/10000... loss: 5.0011... 0.0924 sec/batch
step: 1610/10000... loss: 5.0179... 0.0882 sec/batch
step: 1620/10000... loss: 4.9311... 0.0936 sec/batch
step: 1630/10000... loss: 4.9134... 0.0863 sec/batch
step: 1640/10000... loss: 4.9615... 0.0829 sec/batch
step: 1650/10000... loss: 4.9001... 0.0820 sec/batch
step: 1660/10000... loss: 4.9989... 0.0915 sec/batch
step: 1670/10000... loss: 4.9454... 0.0857 sec/batch
step: 1680/10000... loss: 4.8506... 0.0862 sec/batch
step: 1690/10000... loss: 4.8397... 0.0878 sec/batch
step: 1700/10000... loss: 4.8330... 0.1092 sec/batch
step: 1710/10000... loss: 5.0012... 0.0870 sec/batch
step: 1720/10000... loss: 4.9794... 0.0840 sec/batch
step: 1730/10000... loss: 4.8621... 0.0929 sec/batch
step: 1740/10000... loss: 4.8816... 0.1013 sec/batch
step: 1750/10000... loss: 4.7770... 0.0971 sec/batch
step: 1760/10000... loss: 4.8436... 0.0896 sec/batch
step: 1770/10000... loss: 4.8621... 0.0983 sec/batch
step: 1780/10000... loss: 4.9439... 0.0925 sec/batch
step: 1790/10000... loss: 4.9065... 0.0939 sec/batch
step: 1800/10000... loss: 5.0004... 0.0924 sec/batch
step: 1810/10000... loss: 4.8106... 0.0826 sec/batch
step: 1820/10000... loss: 4.8281... 0.1210 sec/batch
step: 1830/10000... loss: 4.9824... 0.0922 sec/batch
step: 1840/10000... loss: 4.8363... 0.0939 sec/batch
step: 1850/10000... loss: 4.8541... 0.0881 sec/batch
step: 1860/10000... loss: 4.9124... 0.1033 sec/batch
step: 1870/10000... loss: 4.9806... 0.0903 sec/batch
step: 1880/10000... loss: 4.9092... 0.0877 sec/batch
step: 1890/10000... loss: 4.9126... 0.0880 sec/batch
step: 1900/10000... loss: 4.8481... 0.1003 sec/batch
step: 1910/10000... loss: 4.9412... 0.0908 sec/batch
step: 1920/10000... loss: 4.9901... 0.0890 sec/batch
step: 1930/10000... loss: 4.8934... 0.0968 sec/batch
step: 1940/10000... loss: 4.9002... 0.1117 sec/batch
step: 1950/10000... loss: 4.7907... 0.0878 sec/batch
step: 1960/10000... loss: 4.8661... 0.0909 sec/batch
step: 1970/10000... loss: 4.6941... 0.0918 sec/batch
step: 1980/10000... loss: 4.8521... 0.1306 sec/batch
step: 1990/10000... loss: 4.7692... 0.0950 sec/batch
step: 2000/10000... loss: 4.8366... 0.0970 sec/batch
step: 2010/10000... loss: 4.8439... 0.0889 sec/batch
step: 2020/10000... loss: 4.8434... 0.0955 sec/batch
step: 2030/10000... loss: 4.8258... 0.0959 sec/batch
step: 2040/10000... loss: 4.8247... 0.0896 sec/batch
step: 2050/10000... loss: 4.8893... 0.0884 sec/batch
step: 2060/10000... loss: 4.8378... 0.0901 sec/batch
step: 2070/10000... loss: 4.8493... 0.0883 sec/batch
step: 2080/10000... loss: 4.8377... 0.0925 sec/batch
step: 2090/10000... loss: 4.8977... 0.0903 sec/batch
step: 2100/10000... loss: 4.7552... 0.0909 sec/batch
step: 2110/10000... loss: 4.7722... 0.0910 sec/batch
step: 2120/10000... loss: 4.7337... 0.0953 sec/batch
step: 2130/10000... loss: 4.9939... 0.0872 sec/batch
step: 2140/10000... loss: 4.9350... 0.0880 sec/batch
step: 2150/10000... loss: 4.9199... 0.0886 sec/batch
step: 2160/10000... loss: 4.7667... 0.0947 sec/batch
step: 2170/10000... loss: 4.6352... 0.0955 sec/batch
step: 2180/10000... loss: 4.8275... 0.1007 sec/batch
step: 2190/10000... loss: 4.8438... 0.0906 sec/batch
step: 2200/10000... loss: 4.7311... 0.1000 sec/batch
step: 2210/10000... loss: 4.9158... 0.0966 sec/batch
step: 2220/10000... loss: 4.8483... 0.0963 sec/batch
step: 2230/10000... loss: 4.8412... 0.0984 sec/batch
step: 2240/10000... loss: 4.8208... 0.0947 sec/batch
step: 2250/10000... loss: 4.6997... 0.0870 sec/batch
step: 2260/10000... loss: 4.8333... 0.0911 sec/batch
step: 2270/10000... loss: 4.8537... 0.0931 sec/batch
step: 2280/10000... loss: 4.7533... 0.0927 sec/batch
step: 2290/10000... loss: 4.9129... 0.0897 sec/batch
step: 2300/10000... loss: 4.7394... 0.0892 sec/batch
step: 2310/10000... loss: 4.7632... 0.0933 sec/batch
step: 2320/10000... loss: 4.8065... 0.0889 sec/batch
step: 2330/10000... loss: 4.8986... 0.0872 sec/batch
step: 2340/10000... loss: 4.8668... 0.0902 sec/batch
step: 2350/10000... loss: 4.8716... 0.0896 sec/batch
step: 2360/10000... loss: 4.8986... 0.0862 sec/batch
step: 2370/10000... loss: 4.9208... 0.0931 sec/batch
step: 2380/10000... loss: 4.8216... 0.0894 sec/batch
step: 2390/10000... loss: 4.9293... 0.0955 sec/batch
step: 2400/10000... loss: 4.9451... 0.0919 sec/batch
step: 2410/10000... loss: 4.8225... 0.0896 sec/batch
step: 2420/10000... loss: 4.8902... 0.0937 sec/batch
step: 2430/10000... loss: 4.8697... 0.0900 sec/batch
step: 2440/10000... loss: 4.9410... 0.0879 sec/batch
step: 2450/10000... loss: 4.9345... 0.0909 sec/batch
step: 2460/10000... loss: 4.8126... 0.0892 sec/batch
step: 2470/10000... loss: 4.8382... 0.0920 sec/batch
step: 2480/10000... loss: 4.9153... 0.0899 sec/batch
step: 2490/10000... loss: 4.8122... 0.0905 sec/batch
step: 2500/10000... loss: 4.8143... 0.0892 sec/batch
step: 2510/10000... loss: 4.9267... 0.0870 sec/batch
step: 2520/10000... loss: 4.8101... 0.0874 sec/batch
step: 2530/10000... loss: 4.6365... 0.0923 sec/batch
step: 2540/10000... loss: 4.7815... 0.0868 sec/batch
step: 2550/10000... loss: 4.8606... 0.0981 sec/batch
step: 2560/10000... loss: 4.8211... 0.0961 sec/batch
step: 2570/10000... loss: 4.7558... 0.0869 sec/batch
step: 2580/10000... loss: 4.8584... 0.0891 sec/batch
step: 2590/10000... loss: 4.8609... 0.1121 sec/batch
step: 2600/10000... loss: 4.7052... 0.0850 sec/batch
step: 2610/10000... loss: 4.7992... 0.0893 sec/batch
step: 2620/10000... loss: 4.9708... 0.0865 sec/batch
step: 2630/10000... loss: 4.7953... 0.1047 sec/batch
step: 2640/10000... loss: 4.8165... 0.0982 sec/batch
step: 2650/10000... loss: 4.8085... 0.0899 sec/batch
step: 2660/10000... loss: 4.9473... 0.0962 sec/batch
step: 2670/10000... loss: 4.8989... 0.1047 sec/batch
step: 2680/10000... loss: 4.9751... 0.0881 sec/batch
step: 2690/10000... loss: 4.9838... 0.0922 sec/batch
step: 2700/10000... loss: 4.8516... 0.0891 sec/batch
step: 2710/10000... loss: 4.8541... 0.0862 sec/batch
step: 2720/10000... loss: 4.8851... 0.0940 sec/batch
step: 2730/10000... loss: 4.9025... 0.0918 sec/batch
step: 2740/10000... loss: 4.8846... 0.0894 sec/batch
step: 2750/10000... loss: 4.7963... 0.0895 sec/batch
step: 2760/10000... loss: 4.8824... 0.0933 sec/batch
step: 2770/10000... loss: 4.8344... 0.0887 sec/batch
step: 2780/10000... loss: 4.8920... 0.0934 sec/batch
step: 2790/10000... loss: 4.9805... 0.0964 sec/batch
step: 2800/10000... loss: 4.7256... 0.0853 sec/batch
step: 2810/10000... loss: 4.8118... 0.0916 sec/batch
step: 2820/10000... loss: 4.7538... 0.0919 sec/batch
step: 2830/10000... loss: 4.8101... 0.1163 sec/batch
step: 2840/10000... loss: 4.8542... 0.0883 sec/batch
step: 2850/10000... loss: 4.7256... 0.0880 sec/batch
step: 2860/10000... loss: 4.8227... 0.0861 sec/batch
step: 2870/10000... loss: 4.7680... 0.0900 sec/batch
step: 2880/10000... loss: 4.7295... 0.0856 sec/batch
step: 2890/10000... loss: 4.8291... 0.0991 sec/batch
step: 2900/10000... loss: 4.8385... 0.0933 sec/batch
step: 2910/10000... loss: 4.7957... 0.0909 sec/batch
step: 2920/10000... loss: 4.7721... 0.1092 sec/batch
step: 2930/10000... loss: 4.6331... 0.1041 sec/batch
step: 2940/10000... loss: 4.8727... 0.1032 sec/batch
step: 2950/10000... loss: 4.7283... 0.0943 sec/batch
step: 2960/10000... loss: 4.8422... 0.1089 sec/batch
step: 2970/10000... loss: 4.7778... 0.0866 sec/batch
step: 2980/10000... loss: 4.7870... 0.0948 sec/batch
step: 2990/10000... loss: 4.6794... 0.0927 sec/batch
step: 3000/10000... loss: 4.7577... 0.1004 sec/batch
step: 3010/10000... loss: 4.6356... 0.1068 sec/batch
step: 3020/10000... loss: 4.7878... 0.0859 sec/batch
step: 3030/10000... loss: 4.7142... 0.0990 sec/batch
step: 3040/10000... loss: 4.8075... 0.0917 sec/batch
step: 3050/10000... loss: 4.7504... 0.0907 sec/batch
step: 3060/10000... loss: 4.8520... 0.0856 sec/batch
step: 3070/10000... loss: 4.8071... 0.0965 sec/batch
step: 3080/10000... loss: 4.7467... 0.0961 sec/batch
step: 3090/10000... loss: 4.7757... 0.0900 sec/batch
step: 3100/10000... loss: 4.7617... 0.0969 sec/batch
step: 3110/10000... loss: 4.7982... 0.0936 sec/batch
step: 3120/10000... loss: 4.7856... 0.0952 sec/batch
step: 3130/10000... loss: 4.8547... 0.0926 sec/batch
step: 3140/10000... loss: 4.8235... 0.0892 sec/batch
step: 3150/10000... loss: 4.6735... 0.0882 sec/batch
step: 3160/10000... loss: 4.6402... 0.0922 sec/batch
step: 3170/10000... loss: 4.7454... 0.0880 sec/batch
step: 3180/10000... loss: 4.7067... 0.0863 sec/batch
step: 3190/10000... loss: 4.7166... 0.0894 sec/batch
step: 3200/10000... loss: 4.7413... 0.0931 sec/batch
step: 3210/10000... loss: 4.7489... 0.0884 sec/batch
step: 3220/10000... loss: 4.7947... 0.0915 sec/batch
step: 3230/10000... loss: 4.8397... 0.0901 sec/batch
step: 3240/10000... loss: 4.6653... 0.0940 sec/batch
step: 3250/10000... loss: 4.7558... 0.0953 sec/batch
step: 3260/10000... loss: 4.8765... 0.0886 sec/batch
step: 3270/10000... loss: 4.7738... 0.0874 sec/batch
step: 3280/10000... loss: 4.6573... 0.0926 sec/batch
step: 3290/10000... loss: 4.7940... 0.0843 sec/batch
step: 3300/10000... loss: 4.8221... 0.0903 sec/batch
step: 3310/10000... loss: 4.6640... 0.0876 sec/batch
step: 3320/10000... loss: 4.6279... 0.0915 sec/batch
step: 3330/10000... loss: 4.6867... 0.0882 sec/batch
step: 3340/10000... loss: 4.7440... 0.0866 sec/batch
step: 3350/10000... loss: 4.6534... 0.0866 sec/batch
step: 3360/10000... loss: 4.8727... 0.0930 sec/batch
step: 3370/10000... loss: 4.7668... 0.0857 sec/batch
step: 3380/10000... loss: 4.6585... 0.0996 sec/batch
step: 3390/10000... loss: 4.6583... 0.0899 sec/batch
step: 3400/10000... loss: 4.8205... 0.0837 sec/batch
step: 3410/10000... loss: 4.7868... 0.0918 sec/batch
step: 3420/10000... loss: 4.7547... 0.0827 sec/batch
step: 3430/10000... loss: 4.6834... 0.0840 sec/batch
step: 3440/10000... loss: 4.7170... 0.0863 sec/batch
step: 3450/10000... loss: 4.7669... 0.0870 sec/batch
step: 3460/10000... loss: 4.7696... 0.0895 sec/batch
step: 3470/10000... loss: 4.7673... 0.0923 sec/batch
step: 3480/10000... loss: 4.6205... 0.0888 sec/batch
step: 3490/10000... loss: 4.8415... 0.0846 sec/batch
step: 3500/10000... loss: 4.8639... 0.0921 sec/batch
step: 3510/10000... loss: 4.8622... 0.0915 sec/batch
step: 3520/10000... loss: 4.7772... 0.0918 sec/batch
step: 3530/10000... loss: 4.7766... 0.0926 sec/batch
step: 3540/10000... loss: 4.7278... 0.0966 sec/batch
step: 3550/10000... loss: 4.9317... 0.0867 sec/batch
step: 3560/10000... loss: 4.8605... 0.0962 sec/batch
step: 3570/10000... loss: 4.8021... 0.0901 sec/batch
step: 3580/10000... loss: 4.7813... 0.0885 sec/batch
step: 3590/10000... loss: 4.7355... 0.0912 sec/batch
step: 3600/10000... loss: 4.7202... 0.0926 sec/batch
step: 3610/10000... loss: 4.6979... 0.0882 sec/batch
step: 3620/10000... loss: 4.6886... 0.0915 sec/batch
step: 3630/10000... loss: 4.7133... 0.0912 sec/batch
step: 3640/10000... loss: 4.7579... 0.0918 sec/batch
step: 3650/10000... loss: 4.7951... 0.0886 sec/batch
step: 3660/10000... loss: 4.7677... 0.0860 sec/batch
step: 3670/10000... loss: 4.7415... 0.0999 sec/batch
step: 3680/10000... loss: 4.8299... 0.0882 sec/batch
step: 3690/10000... loss: 4.7391... 0.0896 sec/batch
step: 3700/10000... loss: 4.7723... 0.0877 sec/batch
step: 3710/10000... loss: 4.8157... 0.0871 sec/batch
step: 3720/10000... loss: 4.8106... 0.0878 sec/batch
step: 3730/10000... loss: 4.7024... 0.0941 sec/batch
step: 3740/10000... loss: 4.6619... 0.0886 sec/batch
step: 3750/10000... loss: 4.7362... 0.0902 sec/batch
step: 3760/10000... loss: 4.6879... 0.0899 sec/batch
step: 3770/10000... loss: 4.7354... 0.0875 sec/batch
step: 3780/10000... loss: 4.8969... 0.0887 sec/batch
step: 3790/10000... loss: 4.8294... 0.1003 sec/batch
step: 3800/10000... loss: 4.6909... 0.0890 sec/batch
step: 3810/10000... loss: 4.9102... 0.0973 sec/batch
step: 3820/10000... loss: 4.7756... 0.1031 sec/batch
step: 3830/10000... loss: 4.6378... 0.0899 sec/batch
step: 3840/10000... loss: 4.7842... 0.0960 sec/batch
step: 3850/10000... loss: 4.7560... 0.0899 sec/batch
step: 3860/10000... loss: 4.7347... 0.0981 sec/batch
step: 3870/10000... loss: 4.8285... 0.0895 sec/batch
step: 3880/10000... loss: 4.7607... 0.0920 sec/batch
step: 3890/10000... loss: 4.7989... 0.0987 sec/batch
step: 3900/10000... loss: 4.8369... 0.0894 sec/batch
step: 3910/10000... loss: 4.7346... 0.0999 sec/batch
step: 3920/10000... loss: 4.8247... 0.0918 sec/batch
step: 3930/10000... loss: 4.7967... 0.0899 sec/batch
step: 3940/10000... loss: 4.7688... 0.1027 sec/batch
step: 3950/10000... loss: 4.7996... 0.1058 sec/batch
step: 3960/10000... loss: 4.7448... 0.1031 sec/batch
step: 3970/10000... loss: 4.6292... 0.0876 sec/batch
step: 3980/10000... loss: 4.7492... 0.0932 sec/batch
step: 3990/10000... loss: 4.7180... 0.1228 sec/batch
step: 4000/10000... loss: 4.7951... 0.0977 sec/batch
step: 4010/10000... loss: 4.6747... 0.0913 sec/batch
step: 4020/10000... loss: 4.5647... 0.1014 sec/batch
step: 4030/10000... loss: 4.7488... 0.0872 sec/batch
step: 4040/10000... loss: 4.8563... 0.0871 sec/batch
step: 4050/10000... loss: 4.6390... 0.0861 sec/batch
step: 4060/10000... loss: 4.6330... 0.0884 sec/batch
step: 4070/10000... loss: 4.8057... 0.0967 sec/batch
step: 4080/10000... loss: 4.7564... 0.1023 sec/batch
step: 4090/10000... loss: 4.7210... 0.0903 sec/batch
step: 4100/10000... loss: 4.6307... 0.0956 sec/batch
step: 4110/10000... loss: 4.6292... 0.0900 sec/batch
step: 4120/10000... loss: 4.7525... 0.1176 sec/batch
step: 4130/10000... loss: 4.5470... 0.0926 sec/batch
step: 4140/10000... loss: 4.8272... 0.0950 sec/batch
step: 4150/10000... loss: 4.7804... 0.0955 sec/batch
step: 4160/10000... loss: 4.6830... 0.0922 sec/batch
step: 4170/10000... loss: 4.7192... 0.0849 sec/batch
step: 4180/10000... loss: 4.7815... 0.1003 sec/batch
step: 4190/10000... loss: 4.7570... 0.0876 sec/batch
step: 4200/10000... loss: 4.6008... 0.0877 sec/batch
step: 4210/10000... loss: 4.7728... 0.0940 sec/batch
step: 4220/10000... loss: 4.6273... 0.0935 sec/batch
step: 4230/10000... loss: 4.5675... 0.0893 sec/batch
step: 4240/10000... loss: 4.7880... 0.1010 sec/batch
step: 4250/10000... loss: 4.6212... 0.0859 sec/batch
step: 4260/10000... loss: 4.7433... 0.0921 sec/batch
step: 4270/10000... loss: 4.7738... 0.0911 sec/batch
step: 4280/10000... loss: 4.6091... 0.1075 sec/batch
step: 4290/10000... loss: 4.6246... 0.0925 sec/batch
step: 4300/10000... loss: 4.6371... 0.0859 sec/batch
step: 4310/10000... loss: 4.6741... 0.0936 sec/batch
step: 4320/10000... loss: 4.7098... 0.1090 sec/batch
step: 4330/10000... loss: 4.7672... 0.1035 sec/batch
step: 4340/10000... loss: 4.6954... 0.0943 sec/batch
step: 4350/10000... loss: 4.6830... 0.0919 sec/batch
step: 4360/10000... loss: 4.6687... 0.0910 sec/batch
step: 4370/10000... loss: 4.6437... 0.0900 sec/batch
step: 4380/10000... loss: 4.6904... 0.0898 sec/batch
step: 4390/10000... loss: 4.6427... 0.1031 sec/batch
step: 4400/10000... loss: 4.7107... 0.0971 sec/batch
step: 4410/10000... loss: 4.5591... 0.0909 sec/batch
step: 4420/10000... loss: 4.7560... 0.0916 sec/batch
step: 4430/10000... loss: 4.7472... 0.1028 sec/batch
step: 4440/10000... loss: 4.7109... 0.0911 sec/batch
step: 4450/10000... loss: 4.6011... 0.0905 sec/batch
step: 4460/10000... loss: 4.6392... 0.0895 sec/batch
step: 4470/10000... loss: 4.7166... 0.1053 sec/batch
step: 4480/10000... loss: 4.5903... 0.0876 sec/batch
step: 4490/10000... loss: 4.7669... 0.0903 sec/batch
step: 4500/10000... loss: 4.7892... 0.0878 sec/batch
step: 4510/10000... loss: 4.6093... 0.0887 sec/batch
step: 4520/10000... loss: 4.6985... 0.0857 sec/batch
step: 4530/10000... loss: 4.5797... 0.0977 sec/batch
step: 4540/10000... loss: 4.5700... 0.0892 sec/batch
step: 4550/10000... loss: 4.7514... 0.0870 sec/batch
step: 4560/10000... loss: 4.6938... 0.0877 sec/batch
step: 4570/10000... loss: 4.7540... 0.0944 sec/batch
step: 4580/10000... loss: 4.7458... 0.1007 sec/batch
step: 4590/10000... loss: 4.7689... 0.0900 sec/batch
step: 4600/10000... loss: 4.7276... 0.1066 sec/batch
step: 4610/10000... loss: 4.8688... 0.1022 sec/batch
step: 4620/10000... loss: 4.7893... 0.0998 sec/batch
step: 4630/10000... loss: 4.7246... 0.0856 sec/batch
step: 4640/10000... loss: 4.6459... 0.0909 sec/batch
step: 4650/10000... loss: 4.6672... 0.0947 sec/batch
step: 4660/10000... loss: 4.8118... 0.0982 sec/batch
step: 4670/10000... loss: 4.7602... 0.0880 sec/batch
step: 4680/10000... loss: 4.7734... 0.0945 sec/batch
step: 4690/10000... loss: 4.6469... 0.0889 sec/batch
step: 4700/10000... loss: 4.6087... 0.0979 sec/batch
step: 4710/10000... loss: 4.6783... 0.0904 sec/batch
step: 4720/10000... loss: 4.6200... 0.0898 sec/batch
step: 4730/10000... loss: 4.7148... 0.0892 sec/batch
step: 4740/10000... loss: 4.7254... 0.0956 sec/batch
step: 4750/10000... loss: 4.6386... 0.0905 sec/batch
step: 4760/10000... loss: 4.6364... 0.0933 sec/batch
step: 4770/10000... loss: 4.7527... 0.0902 sec/batch
step: 4780/10000... loss: 4.6127... 0.0919 sec/batch
step: 4790/10000... loss: 4.7778... 0.1061 sec/batch
step: 4800/10000... loss: 4.6662... 0.0967 sec/batch
step: 4810/10000... loss: 4.6704... 0.0901 sec/batch
step: 4820/10000... loss: 4.6282... 0.1095 sec/batch
step: 4830/10000... loss: 4.6901... 0.0862 sec/batch
step: 4840/10000... loss: 4.7544... 0.0886 sec/batch
step: 4850/10000... loss: 4.7387... 0.0906 sec/batch
step: 4860/10000... loss: 4.7323... 0.1032 sec/batch
step: 4870/10000... loss: 4.7135... 0.0973 sec/batch
step: 4880/10000... loss: 4.8619... 0.1064 sec/batch
step: 4890/10000... loss: 4.6789... 0.0903 sec/batch
step: 4900/10000... loss: 4.6440... 0.0866 sec/batch
step: 4910/10000... loss: 4.7746... 0.0829 sec/batch
step: 4920/10000... loss: 4.5888... 0.0887 sec/batch
step: 4930/10000... loss: 4.6040... 0.0820 sec/batch
step: 4940/10000... loss: 4.8002... 0.0980 sec/batch
step: 4950/10000... loss: 4.7467... 0.0879 sec/batch
step: 4960/10000... loss: 4.7231... 0.0962 sec/batch
step: 4970/10000... loss: 4.6522... 0.0907 sec/batch
step: 4980/10000... loss: 4.6948... 0.0882 sec/batch
step: 4990/10000... loss: 4.7467... 0.0921 sec/batch
step: 5000/10000... loss: 4.7935... 0.0980 sec/batch
step: 5010/10000... loss: 4.7815... 0.1115 sec/batch
step: 5020/10000... loss: 4.5564... 0.0918 sec/batch
step: 5030/10000... loss: 4.8385... 0.0922 sec/batch
step: 5040/10000... loss: 4.8235... 0.0863 sec/batch
step: 5050/10000... loss: 4.8345... 0.0959 sec/batch
step: 5060/10000... loss: 4.7516... 0.1041 sec/batch
step: 5070/10000... loss: 4.6342... 0.0903 sec/batch
step: 5080/10000... loss: 4.7084... 0.0898 sec/batch
step: 5090/10000... loss: 4.7533... 0.1166 sec/batch
step: 5100/10000... loss: 4.5916... 0.0834 sec/batch
step: 5110/10000... loss: 4.6774... 0.0931 sec/batch
step: 5120/10000... loss: 4.6972... 0.0866 sec/batch
step: 5130/10000... loss: 4.4206... 0.0909 sec/batch
step: 5140/10000... loss: 4.6413... 0.0917 sec/batch
step: 5150/10000... loss: 4.5769... 0.0910 sec/batch
step: 5160/10000... loss: 4.6238... 0.0887 sec/batch
step: 5170/10000... loss: 4.7923... 0.0847 sec/batch
step: 5180/10000... loss: 4.6505... 0.0890 sec/batch
step: 5190/10000... loss: 4.6294... 0.0882 sec/batch
step: 5200/10000... loss: 4.6971... 0.0916 sec/batch
step: 5210/10000... loss: 4.5712... 0.0966 sec/batch
step: 5220/10000... loss: 4.6433... 0.0856 sec/batch
step: 5230/10000... loss: 4.7559... 0.1020 sec/batch
step: 5240/10000... loss: 4.7443... 0.0904 sec/batch
step: 5250/10000... loss: 4.6246... 0.0950 sec/batch
step: 5260/10000... loss: 4.5883... 0.0952 sec/batch
step: 5270/10000... loss: 4.7359... 0.0878 sec/batch
step: 5280/10000... loss: 4.6483... 0.0919 sec/batch
step: 5290/10000... loss: 4.6389... 0.1030 sec/batch
step: 5300/10000... loss: 4.6616... 0.1181 sec/batch
step: 5310/10000... loss: 4.7498... 0.0914 sec/batch
step: 5320/10000... loss: 4.7577... 0.0953 sec/batch
step: 5330/10000... loss: 4.5319... 0.0893 sec/batch
step: 5340/10000... loss: 4.5595... 0.1074 sec/batch
step: 5350/10000... loss: 4.6080... 0.0930 sec/batch
step: 5360/10000... loss: 4.6943... 0.0914 sec/batch
step: 5370/10000... loss: 4.6977... 0.0850 sec/batch
step: 5380/10000... loss: 4.6173... 0.1102 sec/batch
step: 5390/10000... loss: 4.6952... 0.0865 sec/batch
step: 5400/10000... loss: 4.4908... 0.0898 sec/batch
step: 5410/10000... loss: 4.6203... 0.0969 sec/batch
step: 5420/10000... loss: 4.6377... 0.1122 sec/batch
step: 5430/10000... loss: 4.6130... 0.0865 sec/batch
step: 5440/10000... loss: 4.6470... 0.0969 sec/batch
step: 5450/10000... loss: 4.6662... 0.0868 sec/batch
step: 5460/10000... loss: 4.5709... 0.1221 sec/batch
step: 5470/10000... loss: 4.6388... 0.0984 sec/batch
step: 5480/10000... loss: 4.6849... 0.1082 sec/batch
step: 5490/10000... loss: 4.7545... 0.0963 sec/batch
step: 5500/10000... loss: 4.6420... 0.0852 sec/batch
step: 5510/10000... loss: 4.6065... 0.0880 sec/batch
step: 5520/10000... loss: 4.5073... 0.0968 sec/batch
step: 5530/10000... loss: 4.6393... 0.0909 sec/batch
step: 5540/10000... loss: 4.6503... 0.0910 sec/batch
step: 5550/10000... loss: 4.7336... 0.0887 sec/batch
step: 5560/10000... loss: 4.6163... 0.0943 sec/batch
step: 5570/10000... loss: 4.5854... 0.0902 sec/batch
step: 5580/10000... loss: 4.5102... 0.0913 sec/batch
step: 5590/10000... loss: 4.7511... 0.0932 sec/batch
step: 5600/10000... loss: 4.6173... 0.0896 sec/batch
step: 5610/10000... loss: 4.6083... 0.1029 sec/batch
step: 5620/10000... loss: 4.6656... 0.0987 sec/batch
step: 5630/10000... loss: 4.7015... 0.0968 sec/batch
step: 5640/10000... loss: 4.5566... 0.0914 sec/batch
step: 5650/10000... loss: 4.5943... 0.0907 sec/batch
step: 5660/10000... loss: 4.6382... 0.0873 sec/batch
step: 5670/10000... loss: 4.7002... 0.0894 sec/batch
step: 5680/10000... loss: 4.6437... 0.0911 sec/batch
step: 5690/10000... loss: 4.7324... 0.0962 sec/batch
step: 5700/10000... loss: 4.6028... 0.0939 sec/batch
step: 5710/10000... loss: 4.6851... 0.0929 sec/batch
step: 5720/10000... loss: 4.6200... 0.0869 sec/batch
step: 5730/10000... loss: 4.7576... 0.0920 sec/batch
step: 5740/10000... loss: 4.7400... 0.0889 sec/batch
step: 5750/10000... loss: 4.8400... 0.1121 sec/batch
step: 5760/10000... loss: 4.6998... 0.0989 sec/batch
step: 5770/10000... loss: 4.6979... 0.0908 sec/batch
step: 5780/10000... loss: 4.6627... 0.0895 sec/batch
step: 5790/10000... loss: 4.6360... 0.1020 sec/batch
step: 5800/10000... loss: 4.6108... 0.0928 sec/batch
step: 5810/10000... loss: 4.6418... 0.0840 sec/batch
step: 5820/10000... loss: 4.6132... 0.0872 sec/batch
step: 5830/10000... loss: 4.6025... 0.1036 sec/batch
step: 5840/10000... loss: 4.7183... 0.0903 sec/batch
step: 5850/10000... loss: 4.6383... 0.0876 sec/batch
step: 5860/10000... loss: 4.6912... 0.0931 sec/batch
step: 5870/10000... loss: 4.6855... 0.1182 sec/batch
step: 5880/10000... loss: 4.6500... 0.0976 sec/batch
step: 5890/10000... loss: 4.6409... 0.0872 sec/batch
step: 5900/10000... loss: 4.6480... 0.0884 sec/batch
step: 5910/10000... loss: 4.7292... 0.0982 sec/batch
step: 5920/10000... loss: 4.7184... 0.0924 sec/batch
step: 5930/10000... loss: 4.6348... 0.1074 sec/batch
step: 5940/10000... loss: 4.7059... 0.0875 sec/batch
step: 5950/10000... loss: 4.6917... 0.0936 sec/batch
step: 5960/10000... loss: 4.6176... 0.0922 sec/batch
step: 5970/10000... loss: 4.7509... 0.0896 sec/batch
step: 5980/10000... loss: 4.7417... 0.0908 sec/batch
step: 5990/10000... loss: 4.6811... 0.1039 sec/batch
step: 6000/10000... loss: 4.6937... 0.0883 sec/batch
step: 6010/10000... loss: 4.7284... 0.0904 sec/batch
step: 6020/10000... loss: 4.6517... 0.0871 sec/batch
step: 6030/10000... loss: 4.6006... 0.0886 sec/batch
step: 6040/10000... loss: 4.5722... 0.1102 sec/batch
step: 6050/10000... loss: 4.5422... 0.0884 sec/batch
step: 6060/10000... loss: 4.7869... 0.0926 sec/batch
step: 6070/10000... loss: 4.6350... 0.0993 sec/batch
step: 6080/10000... loss: 4.7717... 0.0898 sec/batch
step: 6090/10000... loss: 4.7230... 0.0935 sec/batch
step: 6100/10000... loss: 4.6998... 0.0927 sec/batch
step: 6110/10000... loss: 4.6467... 0.0918 sec/batch
step: 6120/10000... loss: 4.7526... 0.0954 sec/batch
step: 6130/10000... loss: 4.7436... 0.0867 sec/batch
step: 6140/10000... loss: 4.7009... 0.0884 sec/batch
step: 6150/10000... loss: 4.5120... 0.0913 sec/batch
step: 6160/10000... loss: 4.6679... 0.0890 sec/batch
step: 6170/10000... loss: 4.6094... 0.0873 sec/batch
step: 6180/10000... loss: 4.6189... 0.0953 sec/batch
step: 6190/10000... loss: 4.7088... 0.0897 sec/batch
step: 6200/10000... loss: 4.5925... 0.0891 sec/batch
step: 6210/10000... loss: 4.6124... 0.0905 sec/batch
step: 6220/10000... loss: 4.6974... 0.1009 sec/batch
step: 6230/10000... loss: 4.7231... 0.0896 sec/batch
step: 6240/10000... loss: 4.6952... 0.0936 sec/batch
step: 6250/10000... loss: 4.5462... 0.1081 sec/batch
step: 6260/10000... loss: 4.6077... 0.0891 sec/batch
step: 6270/10000... loss: 4.7436... 0.0921 sec/batch
step: 6280/10000... loss: 4.6818... 0.0903 sec/batch
step: 6290/10000... loss: 4.5901... 0.1084 sec/batch
step: 6300/10000... loss: 4.7541... 0.1072 sec/batch
step: 6310/10000... loss: 4.7399... 0.0894 sec/batch
step: 6320/10000... loss: 4.5785... 0.0890 sec/batch
step: 6330/10000... loss: 4.7056... 0.0879 sec/batch
step: 6340/10000... loss: 4.5118... 0.0934 sec/batch
step: 6350/10000... loss: 4.5934... 0.0876 sec/batch
step: 6360/10000... loss: 4.6276... 0.0945 sec/batch
step: 6370/10000... loss: 4.5988... 0.1086 sec/batch
step: 6380/10000... loss: 4.7031... 0.1007 sec/batch
step: 6390/10000... loss: 4.6180... 0.0938 sec/batch
step: 6400/10000... loss: 4.6781... 0.0887 sec/batch
step: 6410/10000... loss: 4.5657... 0.0960 sec/batch
step: 6420/10000... loss: 4.6993... 0.0880 sec/batch
step: 6430/10000... loss: 4.7189... 0.0907 sec/batch
step: 6440/10000... loss: 4.5817... 0.0863 sec/batch
step: 6450/10000... loss: 4.6847... 0.1100 sec/batch
step: 6460/10000... loss: 4.5981... 0.0890 sec/batch
step: 6470/10000... loss: 4.6016... 0.0897 sec/batch
step: 6480/10000... loss: 4.5638... 0.0877 sec/batch
step: 6490/10000... loss: 4.6858... 0.1036 sec/batch
step: 6500/10000... loss: 4.4613... 0.0893 sec/batch
step: 6510/10000... loss: 4.6037... 0.0906 sec/batch
step: 6520/10000... loss: 4.5960... 0.0873 sec/batch
step: 6530/10000... loss: 4.6400... 0.0944 sec/batch
step: 6540/10000... loss: 4.5454... 0.0908 sec/batch
step: 6550/10000... loss: 4.4611... 0.0840 sec/batch
step: 6560/10000... loss: 4.6385... 0.0871 sec/batch
step: 6570/10000... loss: 4.4945... 0.0898 sec/batch
step: 6580/10000... loss: 4.5617... 0.0860 sec/batch
step: 6590/10000... loss: 4.6809... 0.0940 sec/batch
step: 6600/10000... loss: 4.5555... 0.0897 sec/batch
step: 6610/10000... loss: 4.4528... 0.0863 sec/batch
step: 6620/10000... loss: 4.6082... 0.0901 sec/batch
step: 6630/10000... loss: 4.7801... 0.0941 sec/batch
step: 6640/10000... loss: 4.6948... 0.0959 sec/batch
step: 6650/10000... loss: 4.6163... 0.1107 sec/batch
step: 6660/10000... loss: 4.6134... 0.0918 sec/batch
step: 6670/10000... loss: 4.5979... 0.0978 sec/batch
step: 6680/10000... loss: 4.6256... 0.0894 sec/batch
step: 6690/10000... loss: 4.6180... 0.0840 sec/batch
step: 6700/10000... loss: 4.5030... 0.0921 sec/batch
step: 6710/10000... loss: 4.6326... 0.0883 sec/batch
step: 6720/10000... loss: 4.4899... 0.1009 sec/batch
step: 6730/10000... loss: 4.6855... 0.0853 sec/batch
step: 6740/10000... loss: 4.5870... 0.0922 sec/batch
step: 6750/10000... loss: 4.8134... 0.0976 sec/batch
step: 6760/10000... loss: 4.5962... 0.1018 sec/batch
step: 6770/10000... loss: 4.6327... 0.0840 sec/batch
step: 6780/10000... loss: 4.5311... 0.0978 sec/batch
step: 6790/10000... loss: 4.5845... 0.0935 sec/batch
step: 6800/10000... loss: 4.6039... 0.0884 sec/batch
step: 6810/10000... loss: 4.5224... 0.1000 sec/batch
step: 6820/10000... loss: 4.5950... 0.1131 sec/batch
step: 6830/10000... loss: 4.5994... 0.0946 sec/batch
step: 6840/10000... loss: 4.6660... 0.0908 sec/batch
step: 6850/10000... loss: 4.6079... 0.0883 sec/batch
step: 6860/10000... loss: 4.6590... 0.0880 sec/batch
step: 6870/10000... loss: 4.6126... 0.0923 sec/batch
step: 6880/10000... loss: 4.6539... 0.0939 sec/batch
step: 6890/10000... loss: 4.6775... 0.1067 sec/batch
step: 6900/10000... loss: 4.6485... 0.0933 sec/batch
step: 6910/10000... loss: 4.6610... 0.0915 sec/batch
step: 6920/10000... loss: 4.6436... 0.0937 sec/batch
step: 6930/10000... loss: 4.5383... 0.0873 sec/batch
step: 6940/10000... loss: 4.7345... 0.0908 sec/batch
step: 6950/10000... loss: 4.7277... 0.0989 sec/batch
step: 6960/10000... loss: 4.6551... 0.0889 sec/batch
step: 6970/10000... loss: 4.5980... 0.0877 sec/batch
step: 6980/10000... loss: 4.5870... 0.0902 sec/batch
step: 6990/10000... loss: 4.6365... 0.1051 sec/batch
step: 7000/10000... loss: 4.6720... 0.0932 sec/batch
step: 7010/10000... loss: 4.6554... 0.0949 sec/batch
step: 7020/10000... loss: 4.4978... 0.0923 sec/batch
step: 7030/10000... loss: 4.5188... 0.0933 sec/batch
step: 7040/10000... loss: 4.7094... 0.0964 sec/batch
step: 7050/10000... loss: 4.6812... 0.0973 sec/batch
step: 7060/10000... loss: 4.5269... 0.1001 sec/batch
step: 7070/10000... loss: 4.5391... 0.0905 sec/batch
step: 7080/10000... loss: 4.6300... 0.0973 sec/batch
step: 7090/10000... loss: 4.5960... 0.0893 sec/batch
step: 7100/10000... loss: 4.6308... 0.0908 sec/batch
step: 7110/10000... loss: 4.5511... 0.0961 sec/batch
step: 7120/10000... loss: 4.6733... 0.0929 sec/batch
step: 7130/10000... loss: 4.6295... 0.0873 sec/batch
step: 7140/10000... loss: 4.6619... 0.1061 sec/batch
step: 7150/10000... loss: 4.7400... 0.0925 sec/batch
step: 7160/10000... loss: 4.6165... 0.1088 sec/batch
step: 7170/10000... loss: 4.6516... 0.0893 sec/batch
step: 7180/10000... loss: 4.7484... 0.0981 sec/batch
step: 7190/10000... loss: 4.4948... 0.0927 sec/batch
step: 7200/10000... loss: 4.5394... 0.0911 sec/batch
step: 7210/10000... loss: 4.8308... 0.0855 sec/batch
step: 7220/10000... loss: 4.6820... 0.1014 sec/batch
step: 7230/10000... loss: 4.5807... 0.0855 sec/batch
step: 7240/10000... loss: 4.6651... 0.0926 sec/batch
step: 7250/10000... loss: 4.7415... 0.1143 sec/batch
step: 7260/10000... loss: 4.7141... 0.0975 sec/batch
step: 7270/10000... loss: 4.7324... 0.0928 sec/batch
step: 7280/10000... loss: 4.6216... 0.0900 sec/batch
step: 7290/10000... loss: 4.5137... 0.1328 sec/batch
step: 7300/10000... loss: 4.5967... 0.1049 sec/batch
step: 7310/10000... loss: 4.6654... 0.0894 sec/batch
step: 7320/10000... loss: 4.5759... 0.0984 sec/batch
step: 7330/10000... loss: 4.6317... 0.0945 sec/batch
step: 7340/10000... loss: 4.7221... 0.0984 sec/batch
step: 7350/10000... loss: 4.6853... 0.0912 sec/batch
step: 7360/10000... loss: 4.5253... 0.0891 sec/batch
step: 7370/10000... loss: 4.6267... 0.0941 sec/batch
step: 7380/10000... loss: 4.4240... 0.0874 sec/batch
step: 7390/10000... loss: 4.6061... 0.1020 sec/batch
step: 7400/10000... loss: 4.5282... 0.0910 sec/batch
step: 7410/10000... loss: 4.5467... 0.0886 sec/batch
step: 7420/10000... loss: 4.6002... 0.1008 sec/batch
step: 7430/10000... loss: 4.4848... 0.0904 sec/batch
step: 7440/10000... loss: 4.5564... 0.0907 sec/batch
step: 7450/10000... loss: 4.5614... 0.0900 sec/batch
step: 7460/10000... loss: 4.5751... 0.1006 sec/batch
step: 7470/10000... loss: 4.6160... 0.0866 sec/batch
step: 7480/10000... loss: 4.6221... 0.0878 sec/batch
step: 7490/10000... loss: 4.6254... 0.0910 sec/batch
step: 7500/10000... loss: 4.6454... 0.0878 sec/batch
step: 7510/10000... loss: 4.5263... 0.1010 sec/batch
step: 7520/10000... loss: 4.5824... 0.0903 sec/batch
step: 7530/10000... loss: 4.6218... 0.0929 sec/batch
step: 7540/10000... loss: 4.6059... 0.0910 sec/batch
step: 7550/10000... loss: 4.5480... 0.0873 sec/batch
step: 7560/10000... loss: 4.6316... 0.0883 sec/batch
step: 7570/10000... loss: 4.5782... 0.0913 sec/batch
step: 7580/10000... loss: 4.6705... 0.1015 sec/batch
step: 7590/10000... loss: 4.5632... 0.1172 sec/batch
step: 7600/10000... loss: 4.5502... 0.0916 sec/batch
step: 7610/10000... loss: 4.6325... 0.0910 sec/batch
step: 7620/10000... loss: 4.6143... 0.0948 sec/batch
step: 7630/10000... loss: 4.6401... 0.0925 sec/batch
step: 7640/10000... loss: 4.6058... 0.1083 sec/batch
step: 7650/10000... loss: 4.4045... 0.0996 sec/batch
step: 7660/10000... loss: 4.5775... 0.0940 sec/batch
step: 7670/10000... loss: 4.6791... 0.0939 sec/batch
step: 7680/10000... loss: 4.6021... 0.0928 sec/batch
step: 7690/10000... loss: 4.5756... 0.0899 sec/batch
step: 7700/10000... loss: 4.4690... 0.0910 sec/batch
step: 7710/10000... loss: 4.6216... 0.0899 sec/batch
step: 7720/10000... loss: 4.6908... 0.0988 sec/batch
step: 7730/10000... loss: 4.5388... 0.0891 sec/batch
step: 7740/10000... loss: 4.5490... 0.0874 sec/batch
step: 7750/10000... loss: 4.5206... 0.0953 sec/batch
step: 7760/10000... loss: 4.4812... 0.1113 sec/batch
step: 7770/10000... loss: 4.4927... 0.0841 sec/batch
step: 7780/10000... loss: 4.6073... 0.0890 sec/batch
step: 7790/10000... loss: 4.5739... 0.0853 sec/batch
step: 7800/10000... loss: 4.6212... 0.1287 sec/batch
step: 7810/10000... loss: 4.5563... 0.0959 sec/batch
step: 7820/10000... loss: 4.6082... 0.0930 sec/batch
step: 7830/10000... loss: 4.5427... 0.0909 sec/batch
step: 7840/10000... loss: 4.5607... 0.0859 sec/batch
step: 7850/10000... loss: 4.5356... 0.0906 sec/batch
step: 7860/10000... loss: 4.6555... 0.0875 sec/batch
step: 7870/10000... loss: 4.5419... 0.1003 sec/batch
step: 7880/10000... loss: 4.4669... 0.0913 sec/batch
step: 7890/10000... loss: 4.5111... 0.0868 sec/batch
step: 7900/10000... loss: 4.6000... 0.0914 sec/batch
step: 7910/10000... loss: 4.6142... 0.0989 sec/batch
step: 7920/10000... loss: 4.5968... 0.1066 sec/batch
step: 7930/10000... loss: 4.5989... 0.1106 sec/batch
step: 7940/10000... loss: 4.5994... 0.0925 sec/batch
step: 7950/10000... loss: 4.6041... 0.0914 sec/batch
step: 7960/10000... loss: 4.5507... 0.0928 sec/batch
step: 7970/10000... loss: 4.5354... 0.1031 sec/batch
step: 7980/10000... loss: 4.6406... 0.0917 sec/batch
step: 7990/10000... loss: 4.6410... 0.0936 sec/batch
step: 8000/10000... loss: 4.6625... 0.0883 sec/batch
step: 8010/10000... loss: 4.5772... 0.0904 sec/batch
step: 8020/10000... loss: 4.6812... 0.1082 sec/batch
step: 8030/10000... loss: 4.7502... 0.0968 sec/batch
step: 8040/10000... loss: 4.5747... 0.0926 sec/batch
step: 8050/10000... loss: 4.6265... 0.0851 sec/batch
step: 8060/10000... loss: 4.5926... 0.1017 sec/batch
step: 8070/10000... loss: 4.5954... 0.0900 sec/batch
step: 8080/10000... loss: 4.6970... 0.1143 sec/batch
step: 8090/10000... loss: 4.7825... 0.0875 sec/batch
step: 8100/10000... loss: 4.6735... 0.0941 sec/batch
step: 8110/10000... loss: 4.6434... 0.0970 sec/batch
step: 8120/10000... loss: 4.7077... 0.0978 sec/batch
step: 8130/10000... loss: 4.5738... 0.0846 sec/batch
step: 8140/10000... loss: 4.7203... 0.0979 sec/batch
step: 8150/10000... loss: 4.6812... 0.0868 sec/batch
step: 8160/10000... loss: 4.6029... 0.0933 sec/batch
step: 8170/10000... loss: 4.5119... 0.0922 sec/batch
step: 8180/10000... loss: 4.6293... 0.0868 sec/batch
step: 8190/10000... loss: 4.6079... 0.0859 sec/batch
step: 8200/10000... loss: 4.6418... 0.0947 sec/batch
step: 8210/10000... loss: 4.6154... 0.0911 sec/batch
step: 8220/10000... loss: 4.7005... 0.0938 sec/batch
step: 8230/10000... loss: 4.6382... 0.1128 sec/batch
step: 8240/10000... loss: 4.6328... 0.0882 sec/batch
step: 8250/10000... loss: 4.5354... 0.1077 sec/batch
step: 8260/10000... loss: 4.6102... 0.0847 sec/batch
step: 8270/10000... loss: 4.6150... 0.0912 sec/batch
step: 8280/10000... loss: 4.5451... 0.0893 sec/batch
step: 8290/10000... loss: 4.5671... 0.0899 sec/batch
step: 8300/10000... loss: 4.4917... 0.0921 sec/batch
step: 8310/10000... loss: 4.6544... 0.1101 sec/batch
step: 8320/10000... loss: 4.6283... 0.0924 sec/batch
step: 8330/10000... loss: 4.6683... 0.0947 sec/batch
step: 8340/10000... loss: 4.6915... 0.0912 sec/batch
step: 8350/10000... loss: 4.7295... 0.1158 sec/batch
step: 8360/10000... loss: 4.6201... 0.0917 sec/batch
step: 8370/10000... loss: 4.6730... 0.0885 sec/batch
step: 8380/10000... loss: 4.6421... 0.0871 sec/batch
step: 8390/10000... loss: 4.4909... 0.0889 sec/batch
step: 8400/10000... loss: 4.6271... 0.0869 sec/batch
step: 8410/10000... loss: 4.6707... 0.0937 sec/batch
step: 8420/10000... loss: 4.5780... 0.0866 sec/batch
step: 8430/10000... loss: 4.7211... 0.0922 sec/batch
step: 8440/10000... loss: 4.6390... 0.0926 sec/batch
step: 8450/10000... loss: 4.7050... 0.0907 sec/batch
step: 8460/10000... loss: 4.4904... 0.0905 sec/batch
step: 8470/10000... loss: 4.6536... 0.0855 sec/batch
step: 8480/10000... loss: 4.4162... 0.0936 sec/batch
step: 8490/10000... loss: 4.5931... 0.0899 sec/batch
step: 8500/10000... loss: 4.5700... 0.0917 sec/batch
step: 8510/10000... loss: 4.6493... 0.0881 sec/batch
step: 8520/10000... loss: 4.4857... 0.1094 sec/batch
step: 8530/10000... loss: 4.5074... 0.0868 sec/batch
step: 8540/10000... loss: 4.6752... 0.1004 sec/batch
step: 8550/10000... loss: 4.6573... 0.0950 sec/batch
step: 8560/10000... loss: 4.4968... 0.1142 sec/batch
step: 8570/10000... loss: 4.7074... 0.0970 sec/batch
step: 8580/10000... loss: 4.6042... 0.0972 sec/batch
step: 8590/10000... loss: 4.5797... 0.0905 sec/batch
step: 8600/10000... loss: 4.7578... 0.0969 sec/batch
step: 8610/10000... loss: 4.7127... 0.0945 sec/batch
step: 8620/10000... loss: 4.5538... 0.0855 sec/batch
step: 8630/10000... loss: 4.5837... 0.0893 sec/batch
step: 8640/10000... loss: 4.6992... 0.0930 sec/batch
step: 8650/10000... loss: 4.6277... 0.0879 sec/batch
step: 8660/10000... loss: 4.5336... 0.0948 sec/batch
step: 8670/10000... loss: 4.5748... 0.0872 sec/batch
step: 8680/10000... loss: 4.5794... 0.0837 sec/batch
step: 8690/10000... loss: 4.4788... 0.1127 sec/batch
step: 8700/10000... loss: 4.5349... 0.0896 sec/batch
step: 8710/10000... loss: 4.7088... 0.1018 sec/batch
step: 8720/10000... loss: 4.5753... 0.0969 sec/batch
step: 8730/10000... loss: 4.5090... 0.0965 sec/batch
step: 8740/10000... loss: 4.5945... 0.0921 sec/batch
step: 8750/10000... loss: 4.6122... 0.0879 sec/batch
step: 8760/10000... loss: 4.4623... 0.0895 sec/batch
step: 8770/10000... loss: 4.5883... 0.0935 sec/batch
step: 8780/10000... loss: 4.5552... 0.0921 sec/batch
step: 8790/10000... loss: 4.5097... 0.0928 sec/batch
step: 8800/10000... loss: 4.5820... 0.0908 sec/batch
step: 8810/10000... loss: 4.4862... 0.0887 sec/batch
step: 8820/10000... loss: 4.5168... 0.0972 sec/batch
step: 8830/10000... loss: 4.4936... 0.0950 sec/batch
step: 8840/10000... loss: 4.6276... 0.0857 sec/batch
step: 8850/10000... loss: 4.6495... 0.1072 sec/batch
step: 8860/10000... loss: 4.5880... 0.0938 sec/batch
step: 8870/10000... loss: 4.5413... 0.0860 sec/batch
step: 8880/10000... loss: 4.6304... 0.0915 sec/batch
step: 8890/10000... loss: 4.5779... 0.0991 sec/batch
step: 8900/10000... loss: 4.5407... 0.0906 sec/batch
step: 8910/10000... loss: 4.4692... 0.0933 sec/batch
step: 8920/10000... loss: 4.6857... 0.0873 sec/batch
step: 8930/10000... loss: 4.6261... 0.1011 sec/batch
step: 8940/10000... loss: 4.6471... 0.0908 sec/batch
step: 8950/10000... loss: 4.6660... 0.0933 sec/batch
step: 8960/10000... loss: 4.5239... 0.0894 sec/batch
step: 8970/10000... loss: 4.5337... 0.0821 sec/batch
step: 8980/10000... loss: 4.5771... 0.0968 sec/batch
step: 8990/10000... loss: 4.5932... 0.0896 sec/batch
step: 9000/10000... loss: 4.7003... 0.0936 sec/batch
step: 9010/10000... loss: 4.5460... 0.0839 sec/batch
step: 9020/10000... loss: 4.5863... 0.0935 sec/batch
step: 9030/10000... loss: 4.4787... 0.0893 sec/batch
step: 9040/10000... loss: 4.4798... 0.0944 sec/batch
step: 9050/10000... loss: 4.6261... 0.0911 sec/batch
step: 9060/10000... loss: 4.5732... 0.1006 sec/batch
step: 9070/10000... loss: 4.5538... 0.1017 sec/batch
step: 9080/10000... loss: 4.5323... 0.1064 sec/batch
step: 9090/10000... loss: 4.5977... 0.1024 sec/batch
step: 9100/10000... loss: 4.6434... 0.0974 sec/batch
step: 9110/10000... loss: 4.5109... 0.0849 sec/batch
step: 9120/10000... loss: 4.4829... 0.0866 sec/batch
step: 9130/10000... loss: 4.7870... 0.0878 sec/batch
step: 9140/10000... loss: 4.6287... 0.0879 sec/batch
step: 9150/10000... loss: 4.6560... 0.0914 sec/batch
step: 9160/10000... loss: 4.7458... 0.0951 sec/batch
step: 9170/10000... loss: 4.5893... 0.0907 sec/batch
step: 9180/10000... loss: 4.5783... 0.0949 sec/batch
step: 9190/10000... loss: 4.5979... 0.1285 sec/batch
step: 9200/10000... loss: 4.5847... 0.0926 sec/batch
step: 9210/10000... loss: 4.6365... 0.0977 sec/batch
step: 9220/10000... loss: 4.5703... 0.0923 sec/batch
step: 9230/10000... loss: 4.6421... 0.0929 sec/batch
step: 9240/10000... loss: 4.5873... 0.0862 sec/batch
step: 9250/10000... loss: 4.5717... 0.0920 sec/batch
step: 9260/10000... loss: 4.5144... 0.0876 sec/batch
step: 9270/10000... loss: 4.4788... 0.0903 sec/batch
step: 9280/10000... loss: 4.6626... 0.0982 sec/batch
step: 9290/10000... loss: 4.4947... 0.0866 sec/batch
step: 9300/10000... loss: 4.5811... 0.0882 sec/batch
step: 9310/10000... loss: 4.6134... 0.0947 sec/batch
step: 9320/10000... loss: 4.4892... 0.1094 sec/batch
step: 9330/10000... loss: 4.4825... 0.0935 sec/batch
step: 9340/10000... loss: 4.6810... 0.1011 sec/batch
step: 9350/10000... loss: 4.5688... 0.0911 sec/batch
step: 9360/10000... loss: 4.6239... 0.0952 sec/batch
step: 9370/10000... loss: 4.6082... 0.0892 sec/batch
step: 9380/10000... loss: 4.8793... 0.0949 sec/batch
step: 9390/10000... loss: 4.6597... 0.0944 sec/batch
step: 9400/10000... loss: 4.6949... 0.0973 sec/batch
step: 9410/10000... loss: 4.5442... 0.0915 sec/batch
step: 9420/10000... loss: 4.5411... 0.0954 sec/batch
step: 9430/10000... loss: 4.6218... 0.0839 sec/batch
step: 9440/10000... loss: 4.6181... 0.0893 sec/batch
step: 9450/10000... loss: 4.5185... 0.0955 sec/batch
step: 9460/10000... loss: 4.7007... 0.0959 sec/batch
step: 9470/10000... loss: 4.6732... 0.1050 sec/batch
step: 9480/10000... loss: 4.6962... 0.1092 sec/batch
step: 9490/10000... loss: 4.5258... 0.1190 sec/batch
step: 9500/10000... loss: 4.5217... 0.0931 sec/batch
step: 9510/10000... loss: 4.6891... 0.0932 sec/batch
step: 9520/10000... loss: 4.6916... 0.0866 sec/batch
step: 9530/10000... loss: 4.5395... 0.1104 sec/batch
step: 9540/10000... loss: 4.6270... 0.1037 sec/batch
step: 9550/10000... loss: 4.6109... 0.1054 sec/batch
step: 9560/10000... loss: 4.6458... 0.0912 sec/batch
step: 9570/10000... loss: 4.6489... 0.0877 sec/batch
step: 9580/10000... loss: 4.6561... 0.1015 sec/batch
step: 9590/10000... loss: 4.7778... 0.1012 sec/batch
step: 9600/10000... loss: 4.6169... 0.0973 sec/batch
step: 9610/10000... loss: 4.5508... 0.0973 sec/batch
step: 9620/10000... loss: 4.6320... 0.1003 sec/batch
step: 9630/10000... loss: 4.4132... 0.1281 sec/batch
step: 9640/10000... loss: 4.6708... 0.1090 sec/batch
step: 9650/10000... loss: 4.5088... 0.1056 sec/batch
step: 9660/10000... loss: 4.4661... 0.1146 sec/batch
step: 9670/10000... loss: 4.5559... 0.1078 sec/batch
step: 9680/10000... loss: 4.6313... 0.1068 sec/batch
step: 9690/10000... loss: 4.4804... 0.0942 sec/batch
step: 9700/10000... loss: 4.7342... 0.0858 sec/batch
step: 9710/10000... loss: 4.5631... 0.0925 sec/batch
step: 9720/10000... loss: 4.5038... 0.0935 sec/batch
step: 9730/10000... loss: 4.6602... 0.0902 sec/batch
step: 9740/10000... loss: 4.5223... 0.0892 sec/batch
step: 9750/10000... loss: 4.6234... 0.1100 sec/batch
step: 9760/10000... loss: 4.6627... 0.0969 sec/batch
step: 9770/10000... loss: 4.5483... 0.0852 sec/batch
step: 9780/10000... loss: 4.5161... 0.0908 sec/batch
step: 9790/10000... loss: 4.4137... 0.0879 sec/batch
step: 9800/10000... loss: 4.5930... 0.1010 sec/batch
step: 9810/10000... loss: 4.5248... 0.0950 sec/batch
step: 9820/10000... loss: 4.4703... 0.0871 sec/batch
step: 9830/10000... loss: 4.5538... 0.1049 sec/batch
step: 9840/10000... loss: 4.5746... 0.0896 sec/batch
step: 9850/10000... loss: 4.6019... 0.0980 sec/batch
step: 9860/10000... loss: 4.6320... 0.0962 sec/batch
step: 9870/10000... loss: 4.4107... 0.0946 sec/batch
step: 9880/10000... loss: 4.4157... 0.1076 sec/batch
step: 9890/10000... loss: 4.5488... 0.1011 sec/batch
step: 9900/10000... loss: 4.5771... 0.0882 sec/batch
step: 9910/10000... loss: 4.4354... 0.0886 sec/batch
step: 9920/10000... loss: 4.5482... 0.0934 sec/batch
step: 9930/10000... loss: 4.4400... 0.0935 sec/batch
step: 9940/10000... loss: 4.5097... 0.0948 sec/batch
step: 9950/10000... loss: 4.4637... 0.0865 sec/batch
step: 9960/10000... loss: 4.5473... 0.0890 sec/batch
step: 9970/10000... loss: 4.5969... 0.1137 sec/batch
step: 9980/10000... loss: 4.5175... 0.0815 sec/batch
step: 9990/10000... loss: 4.4894... 0.0850 sec/batch
step: 10000/10000... loss: 4.5577... 0.0850 sec/batch
In [10]:
!cd Char-RNN-TensorFlow && python sample.py --use_embedding --converter_path model/poetry/converter.pkl --checkpoint_path model/poetry/ --max_length 300
WARNING:tensorflow:From /content/Char-RNN-TensorFlow/model.py:93: softmax_cross_entropy_with_logits (from tensorflow.python.ops.nn_ops) is deprecated and will be removed in a future version.
Instructions for updating:
Future major versions of TensorFlow will allow gradients to flow
into the labels input on backprop by default.
See @{tf.nn.softmax_cross_entropy_with_logits_v2}.
2018-07-28 14:48:41.426060: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:897] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2018-07-28 14:48:41.426702: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1392] Found device 0 with properties:
name: Tesla K80 major: 3 minor: 7 memoryClockRate(GHz): 0.8235
pciBusID: 0000:00:04.0
totalMemory: 11.17GiB freeMemory: 11.10GiB
2018-07-28 14:48:41.426772: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1471] Adding visible gpu devices: 0
2018-07-28 14:48:41.893570: I tensorflow/core/common_runtime/gpu/gpu_device.cc:952] Device interconnect StreamExecutor with strength 1 edge matrix:
2018-07-28 14:48:41.893656: I tensorflow/core/common_runtime/gpu/gpu_device.cc:958] 0
2018-07-28 14:48:41.893701: I tensorflow/core/common_runtime/gpu/gpu_device.cc:971] 0: N
2018-07-28 14:48:41.894148: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1084] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 10763 MB memory) -> physical GPU (device: 0, name: Tesla K80, pci bus id: 0000:00:04.0, compute capability: 3.7)
Restored from: model/poetry/model-10000
换<unk><unk><unk><unk>。
一夜一相留，何人见清风。
山中不可问，日夕一相知。
一叶无相见，春光自未成。
不堪无事处，不是白云人。
一日一声去，一年山月深。
一年何所识，何计不相寻。
一日无人远，春山在故乡。
一年无限处，不得不成名。
不得无人去，何人见故人。
不知江海上，一处向江边。
一日无归处，孤舟在北城。
山中春色远，月落月中深。
一夜山边寺，春来一夜声。
何时无限路，一日不归山。
不见南山路，相知有一年。
江湖无不得，山水有无心。
白发何时远，南江一径新。
江山何处去，云上故乡来。
不有青青外，何人不见身。
山风一夜雨，云色入秋风。
何处不知客，相逢何不知。
一年多一醉，何事在青山。
一日江城外，春山
In [18]:
!cd Char-RNN-TensorFlow && python sample.py --use_embedding --converter_path model/poetry/converter.pkl --checkpoint_path model/poetry/ --max_length 300
WARNING:tensorflow:From /content/Char-RNN-TensorFlow/model.py:93: softmax_cross_entropy_with_logits (from tensorflow.python.ops.nn_ops) is deprecated and will be removed in a future version.
Instructions for updating:
Future major versions of TensorFlow will allow gradients to flow
into the labels input on backprop by default.
See @{tf.nn.softmax_cross_entropy_with_logits_v2}.
2018-07-28 15:56:49.830940: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:897] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2018-07-28 15:56:49.831640: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1392] Found device 0 with properties:
name: Tesla K80 major: 3 minor: 7 memoryClockRate(GHz): 0.8235
pciBusID: 0000:00:04.0
totalMemory: 11.17GiB freeMemory: 11.10GiB
2018-07-28 15:56:49.831708: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1471] Adding visible gpu devices: 0
2018-07-28 15:56:50.376510: I tensorflow/core/common_runtime/gpu/gpu_device.cc:952] Device interconnect StreamExecutor with strength 1 edge matrix:
2018-07-28 15:56:50.376649: I tensorflow/core/common_runtime/gpu/gpu_device.cc:958] 0
2018-07-28 15:56:50.376680: I tensorflow/core/common_runtime/gpu/gpu_device.cc:971] 0: N
2018-07-28 15:56:50.377229: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1084] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 10763 MB memory) -> physical GPU (device: 0, name: Tesla K80, pci bus id: 0000:00:04.0, compute capability: 3.7)
Restored from: model/poetry/model-10000
换马去人人。
一片不相逢，一年人亦归。
不见江山去，相思有此时。
江湖一日雨，春水一行人。
白发秋风起，春风夜色深。
一年无限路，何事是无穷。
一日秋风起，相逢不可闻。
不堪归路去，何事更无穷。
日暮云山下，秋风月色中。
一时何处见，不见旧城中。
不得青山里，还知白发游。
不知江海去，不得旧人归。
日日山中水，秋风夜不知。
一来无旧事，无事不同人。
不见江南路，相思不见年。
不知人未尽，何处有无穷。
何必长山去，何如此日来。
一年无限路，不觉有时情。
白日秋光远，孤舟日月深。
江山无一日，山月向江边。
何处归山外，无人不得心。
何人见归去，不得在人心。
白发无时远，无时见一声。
山中山下后，月月
In [16]:
!cd Char-RNN-TensorFlow && python sample.py --use_embedding --converter_path model/poetry/converter.pkl --checkpoint_path model/poetry/ --max_length 300
WARNING:tensorflow:From /content/Char-RNN-TensorFlow/model.py:93: softmax_cross_entropy_with_logits (from tensorflow.python.ops.nn_ops) is deprecated and will be removed in a future version.
Instructions for updating:
Future major versions of TensorFlow will allow gradients to flow
into the labels input on backprop by default.
See @{tf.nn.softmax_cross_entropy_with_logits_v2}.
2018-07-28 15:55:51.179846: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:897] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2018-07-28 15:55:51.180428: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1392] Found device 0 with properties:
name: Tesla K80 major: 3 minor: 7 memoryClockRate(GHz): 0.8235
pciBusID: 0000:00:04.0
totalMemory: 11.17GiB freeMemory: 11.10GiB
2018-07-28 15:55:51.180482: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1471] Adding visible gpu devices: 0
2018-07-28 15:55:51.659481: I tensorflow/core/common_runtime/gpu/gpu_device.cc:952] Device interconnect StreamExecutor with strength 1 edge matrix:
2018-07-28 15:55:51.659602: I tensorflow/core/common_runtime/gpu/gpu_device.cc:958] 0
2018-07-28 15:55:51.659647: I tensorflow/core/common_runtime/gpu/gpu_device.cc:971] 0: N
2018-07-28 15:55:51.660185: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1084] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 10763 MB memory) -> physical GPU (device: 0, name: Tesla K80, pci bus id: 0000:00:04.0, compute capability: 3.7)
Restored from: model/poetry/model-10000
巫山上。
不见一时心，相思有相见。
一日无一声，春山不可待。
春风入云树，落日落秋色。
不见江山月，不知无限客。
山风入山月，山影一声发。
春草有山月，孤舟不得人。
一日秋草下，春山春草深。
一时不得事，相见无时心。
春风不可见，春雨不成天。
何处有人事，不知江外来。
一声春草里，白发白头间。
不觉无心事，无心不可归。
江边一夜雨，月落不成风。
不得长生处，何人见此心。
江中一千里，山下不相留。
一夜无人去，何人是故人。
江湖不得别，何处是归期。
一日不得事，何年不可知。
一回江汉水，不见故人稀。
何处有人意，无人不可知。
何时无此事，相见在南风。
何处不可见，何人见此心。
不知山上路，一夜更何
In [11]:
!curl https://gist.githubusercontent.com/utensil/93b65c93364059246b1321e3e0a6e0fe/raw/22e6ac2f03131fe4d78013873350a6febf91bcda/writings.txt > Char-RNN-TensorFlow/data/writings.txt
% Total % Received % Xferd Average Speed Time Time Time Current
Dload Upload Total Spent Left Speed
100 358k 100 358k 0 0 358k 0 0:00:01 --:--:-- 0:00:01 1683k
In [12]:
!head Char-RNN-TensorFlow/data/writings.txt
岩浆
----
时偿积欠缺填勉， 凝途失涩兴阑珊。
奔流方遒黯如没， 趣中汲静思里安。
泳夜
----
In [13]:
!cd Char-RNN-TensorFlow && python train.py --use_embedding True --input_file data/writings.txt --num_steps 80 --name writings --learning_rate 0.005 --num_seqs 32 --num_layers 3 --embedding_size 256 --lstm_size 256 --max_steps 10000
3029
3030
WARNING:tensorflow:From /content/Char-RNN-TensorFlow/model.py:93: softmax_cross_entropy_with_logits (from tensorflow.python.ops.nn_ops) is deprecated and will be removed in a future version.
Instructions for updating:
Future major versions of TensorFlow will allow gradients to flow
into the labels input on backprop by default.
See @{tf.nn.softmax_cross_entropy_with_logits_v2}.
2018-07-28 14:53:04.354325: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:897] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2018-07-28 14:53:04.355069: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1392] Found device 0 with properties:
name: Tesla K80 major: 3 minor: 7 memoryClockRate(GHz): 0.8235
pciBusID: 0000:00:04.0
totalMemory: 11.17GiB freeMemory: 11.10GiB
2018-07-28 14:53:04.355153: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1471] Adding visible gpu devices: 0
2018-07-28 14:53:04.899401: I tensorflow/core/common_runtime/gpu/gpu_device.cc:952] Device interconnect StreamExecutor with strength 1 edge matrix:
2018-07-28 14:53:04.899489: I tensorflow/core/common_runtime/gpu/gpu_device.cc:958] 0
2018-07-28 14:53:04.899521: I tensorflow/core/common_runtime/gpu/gpu_device.cc:971] 0: N
2018-07-28 14:53:04.899881: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1084] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 10763 MB memory) -> physical GPU (device: 0, name: Tesla K80, pci bus id: 0000:00:04.0, compute capability: 3.7)
step: 10/10000... loss: 6.5528... 0.3863 sec/batch
step: 20/10000... loss: 6.3371... 0.4065 sec/batch
step: 30/10000... loss: 6.2721... 0.3938 sec/batch
step: 40/10000... loss: 6.2588... 0.3733 sec/batch
step: 50/10000... loss: 6.2303... 0.3929 sec/batch
step: 60/10000... loss: 6.1394... 0.3654 sec/batch
step: 70/10000... loss: 6.2148... 0.3766 sec/batch
step: 80/10000... loss: 5.9511... 0.4000 sec/batch
step: 90/10000... loss: 5.8576... 0.4187 sec/batch
step: 100/10000... loss: 6.0163... 0.3892 sec/batch
step: 110/10000... loss: 5.8947... 0.3591 sec/batch
step: 120/10000... loss: 5.7140... 0.3653 sec/batch
step: 130/10000... loss: 5.6042... 0.3671 sec/batch
step: 140/10000... loss: 5.7144... 0.3840 sec/batch
step: 150/10000... loss: 5.6692... 0.3783 sec/batch
step: 160/10000... loss: 5.6157... 0.3726 sec/batch
step: 170/10000... loss: 5.6349... 0.3567 sec/batch
step: 180/10000... loss: 5.4433... 0.3522 sec/batch
step: 190/10000... loss: 5.5300... 0.3734 sec/batch
step: 200/10000... loss: 5.6150... 0.3990 sec/batch
step: 210/10000... loss: 5.5543... 0.4073 sec/batch
step: 220/10000... loss: 5.5812... 0.4080 sec/batch
step: 230/10000... loss: 5.4584... 0.3736 sec/batch
step: 240/10000... loss: 5.3706... 0.3523 sec/batch
step: 250/10000... loss: 5.4954... 0.3429 sec/batch
step: 260/10000... loss: 5.4528... 0.3674 sec/batch
step: 270/10000... loss: 5.2056... 0.3546 sec/batch
step: 280/10000... loss: 5.3320... 0.3517 sec/batch
step: 290/10000... loss: 5.2262... 0.3556 sec/batch
step: 300/10000... loss: 5.4526... 0.3542 sec/batch
step: 310/10000... loss: 5.2831... 0.3438 sec/batch
step: 320/10000... loss: 5.2162... 0.3576 sec/batch
step: 330/10000... loss: 5.3452... 0.3223 sec/batch
step: 340/10000... loss: 5.1004... 0.3483 sec/batch
step: 350/10000... loss: 5.0863... 0.3444 sec/batch
step: 360/10000... loss: 5.3384... 0.3447 sec/batch
step: 370/10000... loss: 5.2611... 0.3778 sec/batch
step: 380/10000... loss: 5.0635... 0.3699 sec/batch
step: 390/10000... loss: 4.9378... 0.3674 sec/batch
step: 400/10000... loss: 5.0683... 0.3618 sec/batch
step: 410/10000... loss: 5.0456... 0.3767 sec/batch
step: 420/10000... loss: 5.0048... 0.3545 sec/batch
step: 430/10000... loss: 5.0199... 0.3471 sec/batch
step: 440/10000... loss: 4.8375... 0.4050 sec/batch
step: 450/10000... loss: 5.0237... 0.3508 sec/batch
step: 460/10000... loss: 5.0744... 0.3514 sec/batch
step: 470/10000... loss: 5.0378... 0.3434 sec/batch
step: 480/10000... loss: 5.0815... 0.3651 sec/batch
step: 490/10000... loss: 4.9712... 0.4092 sec/batch
step: 500/10000... loss: 4.8911... 0.3411 sec/batch
step: 510/10000... loss: 5.0074... 0.3856 sec/batch
step: 520/10000... loss: 4.9737... 0.3690 sec/batch
step: 530/10000... loss: 4.7998... 0.3811 sec/batch
step: 540/10000... loss: 4.9263... 0.3966 sec/batch
step: 550/10000... loss: 4.8151... 0.3455 sec/batch
step: 560/10000... loss: 5.0504... 0.3417 sec/batch
step: 570/10000... loss: 4.8361... 0.3414 sec/batch
step: 580/10000... loss: 4.8241... 0.3572 sec/batch
step: 590/10000... loss: 4.9783... 0.3591 sec/batch
step: 600/10000... loss: 4.6912... 0.4087 sec/batch
step: 610/10000... loss: 4.7281... 0.3660 sec/batch
step: 620/10000... loss: 4.9890... 0.3504 sec/batch
step: 630/10000... loss: 4.8990... 0.3446 sec/batch
step: 640/10000... loss: 4.6740... 0.3450 sec/batch
step: 650/10000... loss: 4.5718... 0.4262 sec/batch
step: 660/10000... loss: 4.7221... 0.4059 sec/batch
step: 670/10000... loss: 4.7162... 0.3759 sec/batch
step: 680/10000... loss: 4.6300... 0.3825 sec/batch
step: 690/10000... loss: 4.6959... 0.3660 sec/batch
step: 700/10000... loss: 4.5142... 0.3580 sec/batch
step: 710/10000... loss: 4.7169... 0.3574 sec/batch
step: 720/10000... loss: 4.7674... 0.3721 sec/batch
step: 730/10000... loss: 4.7836... 0.3614 sec/batch
step: 740/10000... loss: 4.7416... 0.3548 sec/batch
step: 750/10000... loss: 4.6845... 0.3704 sec/batch
step: 760/10000... loss: 4.6068... 0.3599 sec/batch
step: 770/10000... loss: 4.7097... 0.3488 sec/batch
step: 780/10000... loss: 4.6885... 0.3772 sec/batch
step: 790/10000... loss: 4.5074... 0.3989 sec/batch
step: 800/10000... loss: 4.6378... 0.3447 sec/batch
step: 810/10000... loss: 4.5586... 0.3849 sec/batch
step: 820/10000... loss: 4.8078... 0.3857 sec/batch
step: 830/10000... loss: 4.5871... 0.3871 sec/batch
step: 840/10000... loss: 4.5512... 0.3737 sec/batch
step: 850/10000... loss: 4.6951... 0.3932 sec/batch
step: 860/10000... loss: 4.4475... 0.3406 sec/batch
step: 870/10000... loss: 4.4440... 0.3595 sec/batch
step: 880/10000... loss: 4.7116... 0.3667 sec/batch
step: 890/10000... loss: 4.6447... 0.3437 sec/batch
step: 900/10000... loss: 4.3886... 0.3644 sec/batch
step: 910/10000... loss: 4.3531... 0.3610 sec/batch
step: 920/10000... loss: 4.4721... 0.3507 sec/batch
step: 930/10000... loss: 4.4785... 0.3580 sec/batch
step: 940/10000... loss: 4.4021... 0.3512 sec/batch
step: 950/10000... loss: 4.4578... 0.3834 sec/batch
step: 960/10000... loss: 4.2982... 0.3569 sec/batch
step: 970/10000... loss: 4.5150... 0.3449 sec/batch
step: 980/10000... loss: 4.5141... 0.3851 sec/batch
step: 990/10000... loss: 4.5233... 0.3593 sec/batch
step: 1000/10000... loss: 4.5321... 0.3749 sec/batch
step: 1010/10000... loss: 4.4673... 0.3613 sec/batch
step: 1020/10000... loss: 4.3742... 0.3610 sec/batch
step: 1030/10000... loss: 4.5113... 0.3832 sec/batch
step: 1040/10000... loss: 4.4524... 0.3775 sec/batch
step: 1050/10000... loss: 4.2954... 0.3630 sec/batch
step: 1060/10000... loss: 4.3932... 0.3803 sec/batch
step: 1070/10000... loss: 4.3451... 0.3670 sec/batch
step: 1080/10000... loss: 4.5815... 0.3486 sec/batch
step: 1090/10000... loss: 4.3942... 0.3544 sec/batch
step: 1100/10000... loss: 4.3556... 0.3893 sec/batch
step: 1110/10000... loss: 4.4578... 0.3783 sec/batch
step: 1120/10000... loss: 4.2363... 0.3645 sec/batch
step: 1130/10000... loss: 4.2632... 0.3890 sec/batch
step: 1140/10000... loss: 4.5337... 0.3744 sec/batch
step: 1150/10000... loss: 4.4940... 0.3550 sec/batch
step: 1160/10000... loss: 4.1857... 0.3578 sec/batch
step: 1170/10000... loss: 4.1688... 0.3524 sec/batch
step: 1180/10000... loss: 4.2638... 0.3484 sec/batch
step: 1190/10000... loss: 4.2818... 0.3596 sec/batch
step: 1200/10000... loss: 4.2639... 0.3866 sec/batch
step: 1210/10000... loss: 4.2737... 0.3808 sec/batch
step: 1220/10000... loss: 4.1124... 0.3531 sec/batch
step: 1230/10000... loss: 4.2834... 0.2994 sec/batch
step: 1240/10000... loss: 4.3472... 0.3191 sec/batch
step: 1250/10000... loss: 4.3594... 0.3069 sec/batch
step: 1260/10000... loss: 4.3163... 0.3274 sec/batch
step: 1270/10000... loss: 4.2708... 0.3082 sec/batch
step: 1280/10000... loss: 4.2379... 0.3267 sec/batch
step: 1290/10000... loss: 4.3383... 0.3055 sec/batch
step: 1300/10000... loss: 4.3164... 0.3140 sec/batch
step: 1310/10000... loss: 4.1048... 0.3072 sec/batch
step: 1320/10000... loss: 4.2452... 0.3665 sec/batch
step: 1330/10000... loss: 4.1630... 0.3748 sec/batch
step: 1340/10000... loss: 4.4254... 0.3985 sec/batch
step: 1350/10000... loss: 4.2276... 0.3249 sec/batch
step: 1360/10000... loss: 4.1796... 0.3792 sec/batch
step: 1370/10000... loss: 4.3057... 0.3513 sec/batch
step: 1380/10000... loss: 4.1040... 0.3429 sec/batch
step: 1390/10000... loss: 4.0917... 0.3592 sec/batch
step: 1400/10000... loss: 4.3592... 0.3524 sec/batch
step: 1410/10000... loss: 4.3507... 0.3661 sec/batch
step: 1420/10000... loss: 4.0915... 0.3592 sec/batch
step: 1430/10000... loss: 4.0017... 0.3162 sec/batch
step: 1440/10000... loss: 4.1431... 0.2986 sec/batch
step: 1450/10000... loss: 4.1451... 0.2981 sec/batch
step: 1460/10000... loss: 4.1202... 0.3097 sec/batch
step: 1470/10000... loss: 4.1212... 0.3341 sec/batch
step: 1480/10000... loss: 3.9958... 0.3650 sec/batch
step: 1490/10000... loss: 4.1306... 0.3810 sec/batch
step: 1500/10000... loss: 4.2299... 0.3022 sec/batch
step: 1510/10000... loss: 4.2129... 0.3089 sec/batch
step: 1520/10000... loss: 4.2145... 0.3019 sec/batch
step: 1530/10000... loss: 4.1110... 0.3047 sec/batch
step: 1540/10000... loss: 4.1038... 0.3801 sec/batch
step: 1550/10000... loss: 4.1992... 0.3027 sec/batch
step: 1560/10000... loss: 4.2211... 0.3100 sec/batch
step: 1570/10000... loss: 4.0023... 0.3500 sec/batch
step: 1580/10000... loss: 4.1531... 0.3571 sec/batch
step: 1590/10000... loss: 4.0515... 0.3010 sec/batch
step: 1600/10000... loss: 4.2309... 0.3576 sec/batch
step: 1610/10000... loss: 4.0714... 0.3775 sec/batch
step: 1620/10000... loss: 4.0997... 0.3609 sec/batch
step: 1630/10000... loss: 4.1630... 0.3474 sec/batch
step: 1640/10000... loss: 3.9600... 0.3997 sec/batch
step: 1650/10000... loss: 3.9851... 0.4445 sec/batch
step: 1660/10000... loss: 4.2264... 0.3753 sec/batch
step: 1670/10000... loss: 4.1864... 0.4448 sec/batch
step: 1680/10000... loss: 3.9220... 0.4677 sec/batch
step: 1690/10000... loss: 3.9114... 0.3652 sec/batch
step: 1700/10000... loss: 4.0577... 0.2944 sec/batch
step: 1710/10000... loss: 4.0330... 0.3053 sec/batch
step: 1720/10000... loss: 3.9829... 0.3588 sec/batch
step: 1730/10000... loss: 4.0232... 0.4638 sec/batch
step: 1740/10000... loss: 3.8819... 0.3460 sec/batch
step: 1750/10000... loss: 4.0287... 0.3735 sec/batch
step: 1760/10000... loss: 4.0872... 0.3019 sec/batch
step: 1770/10000... loss: 4.1200... 0.3023 sec/batch
step: 1780/10000... loss: 4.0248... 0.3892 sec/batch
step: 1790/10000... loss: 4.0382... 0.3671 sec/batch
step: 1800/10000... loss: 3.9977... 0.3755 sec/batch
step: 1810/10000... loss: 4.1087... 0.4055 sec/batch
step: 1820/10000... loss: 4.0827... 0.3945 sec/batch
step: 1830/10000... loss: 3.9056... 0.3792 sec/batch
step: 1840/10000... loss: 4.0186... 0.2970 sec/batch
step: 1850/10000... loss: 3.9244... 0.3239 sec/batch
step: 1860/10000... loss: 4.1570... 0.3023 sec/batch
step: 1870/10000... loss: 4.0071... 0.3489 sec/batch
step: 1880/10000... loss: 3.9597... 0.3634 sec/batch
step: 1890/10000... loss: 4.0406... 0.4127 sec/batch
step: 1900/10000... loss: 3.8083... 0.3980 sec/batch
step: 1910/10000... loss: 3.9085... 0.3053 sec/batch
step: 1920/10000... loss: 4.1244... 0.3241 sec/batch
step: 1930/10000... loss: 4.0733... 0.3163 sec/batch
step: 1940/10000... loss: 3.7996... 0.3626 sec/batch
step: 1950/10000... loss: 3.8524... 0.3672 sec/batch
step: 1960/10000... loss: 3.9197... 0.3642 sec/batch
step: 1970/10000... loss: 3.9330... 0.3020 sec/batch
step: 1980/10000... loss: 3.8782... 0.2979 sec/batch
step: 1990/10000... loss: 3.9142... 0.3494 sec/batch
step: 2000/10000... loss: 3.7746... 0.3740 sec/batch
step: 2010/10000... loss: 3.8967... 0.3657 sec/batch
step: 2020/10000... loss: 4.0239... 0.3583 sec/batch
step: 2030/10000... loss: 4.0038... 0.3126 sec/batch
step: 2040/10000... loss: 3.9852... 0.3175 sec/batch
step: 2050/10000... loss: 3.9709... 0.2990 sec/batch
step: 2060/10000... loss: 3.9172... 0.3599 sec/batch
step: 2070/10000... loss: 3.9883... 0.3431 sec/batch
step: 2080/10000... loss: 3.9866... 0.4040 sec/batch
step: 2090/10000... loss: 3.8017... 0.3658 sec/batch
step: 2100/10000... loss: 3.9495... 0.3763 sec/batch
step: 2110/10000... loss: 3.8285... 0.4149 sec/batch
step: 2120/10000... loss: 4.0369... 0.3167 sec/batch
step: 2130/10000... loss: 3.8738... 0.3618 sec/batch
step: 2140/10000... loss: 3.8720... 0.3908 sec/batch
step: 2150/10000... loss: 3.9392... 0.3616 sec/batch
step: 2160/10000... loss: 3.7432... 0.3813 sec/batch
step: 2170/10000... loss: 3.7732... 0.3444 sec/batch
step: 2180/10000... loss: 4.0171... 0.3105 sec/batch
step: 2190/10000... loss: 4.0009... 0.3106 sec/batch
step: 2200/10000... loss: 3.7382... 0.3026 sec/batch
step: 2210/10000... loss: 3.7511... 0.3011 sec/batch
step: 2220/10000... loss: 3.8172... 0.3090 sec/batch
step: 2230/10000... loss: 3.8597... 0.3025 sec/batch
step: 2240/10000... loss: 3.8610... 0.3651 sec/batch
step: 2250/10000... loss: 3.8791... 0.3555 sec/batch
step: 2260/10000... loss: 3.7018... 0.3596 sec/batch
step: 2270/10000... loss: 3.8272... 0.3518 sec/batch
step: 2280/10000... loss: 3.9058... 0.3429 sec/batch
step: 2290/10000... loss: 3.9462... 0.3417 sec/batch
step: 2300/10000... loss: 3.8936... 0.3608 sec/batch
step: 2310/10000... loss: 3.8373... 0.3620 sec/batch
step: 2320/10000... loss: 3.8111... 0.3495 sec/batch
step: 2330/10000... loss: 3.9381... 0.4245 sec/batch
step: 2340/10000... loss: 3.9296... 0.3944 sec/batch
step: 2350/10000... loss: 3.7179... 0.3590 sec/batch
step: 2360/10000... loss: 3.8697... 0.3692 sec/batch
step: 2370/10000... loss: 3.7543... 0.3104 sec/batch
step: 2380/10000... loss: 3.9665... 0.3199 sec/batch
step: 2390/10000... loss: 3.8271... 0.3485 sec/batch
step: 2400/10000... loss: 3.8107... 0.3647 sec/batch
step: 2410/10000... loss: 3.8799... 0.3515 sec/batch
step: 2420/10000... loss: 3.6902... 0.4303 sec/batch
step: 2430/10000... loss: 3.7210... 0.3517 sec/batch
step: 2440/10000... loss: 3.9415... 0.3511 sec/batch
step: 2450/10000... loss: 3.9181... 0.3927 sec/batch
step: 2460/10000... loss: 3.6478... 0.3632 sec/batch
step: 2470/10000... loss: 3.6882... 0.4018 sec/batch
step: 2480/10000... loss: 3.7603... 0.4120 sec/batch
step: 2490/10000... loss: 3.7769... 0.3046 sec/batch
step: 2500/10000... loss: 3.7848... 0.3137 sec/batch
step: 2510/10000... loss: 3.7864... 0.3058 sec/batch
step: 2520/10000... loss: 3.6097... 0.3269 sec/batch
step: 2530/10000... loss: 3.7772... 0.3768 sec/batch
step: 2540/10000... loss: 3.8392... 0.3748 sec/batch
step: 2550/10000... loss: 3.8417... 0.3802 sec/batch
step: 2560/10000... loss: 3.8051... 0.3468 sec/batch
step: 2570/10000... loss: 3.7969... 0.3525 sec/batch
step: 2580/10000... loss: 3.7454... 0.3514 sec/batch
step: 2590/10000... loss: 3.8706... 0.3702 sec/batch
step: 2600/10000... loss: 3.8572... 0.3506 sec/batch
step: 2610/10000... loss: 3.6365... 0.3446 sec/batch
step: 2620/10000... loss: 3.7588... 0.3580 sec/batch
step: 2630/10000... loss: 3.7348... 0.3849 sec/batch
step: 2640/10000... loss: 3.9317... 0.3721 sec/batch
step: 2650/10000... loss: 3.7590... 0.3727 sec/batch
step: 2660/10000... loss: 3.8117... 0.3596 sec/batch
step: 2670/10000... loss: 3.8086... 0.3677 sec/batch
step: 2680/10000... loss: 3.6724... 0.3515 sec/batch
step: 2690/10000... loss: 3.6460... 0.4057 sec/batch
step: 2700/10000... loss: 3.8833... 0.3526 sec/batch
step: 2710/10000... loss: 3.8245... 0.3446 sec/batch
step: 2720/10000... loss: 3.6061... 0.3611 sec/batch
step: 2730/10000... loss: 3.6781... 0.3548 sec/batch
step: 2740/10000... loss: 3.6996... 0.3743 sec/batch
step: 2750/10000... loss: 3.7530... 0.3413 sec/batch
step: 2760/10000... loss: 3.7125... 0.3415 sec/batch
step: 2770/10000... loss: 3.7317... 0.3670 sec/batch
step: 2780/10000... loss: 3.5952... 0.3744 sec/batch
step: 2790/10000... loss: 3.7537... 0.3698 sec/batch
step: 2800/10000... loss: 3.7948... 0.3163 sec/batch
step: 2810/10000... loss: 3.8052... 0.2987 sec/batch
step: 2820/10000... loss: 3.7550... 0.3219 sec/batch
step: 2830/10000... loss: 3.7676... 0.3730 sec/batch
step: 2840/10000... loss: 3.7447... 0.3630 sec/batch
step: 2850/10000... loss: 3.8129... 0.3616 sec/batch
step: 2860/10000... loss: 3.7836... 0.3631 sec/batch
step: 2870/10000... loss: 3.5740... 0.3651 sec/batch
step: 2880/10000... loss: 3.7347... 0.3873 sec/batch
step: 2890/10000... loss: 3.6798... 0.3841 sec/batch
step: 2900/10000... loss: 3.8112... 0.3618 sec/batch
step: 2910/10000... loss: 3.7194... 0.3025 sec/batch
step: 2920/10000... loss: 3.7442... 0.3297 sec/batch
step: 2930/10000... loss: 3.7698... 0.3030 sec/batch
step: 2940/10000... loss: 3.6009... 0.2960 sec/batch
step: 2950/10000... loss: 3.6209... 0.3088 sec/batch
step: 2960/10000... loss: 3.8759... 0.3034 sec/batch
step: 2970/10000... loss: 3.8500... 0.3003 sec/batch
step: 2980/10000... loss: 3.5834... 0.3080 sec/batch
step: 2990/10000... loss: 3.6128... 0.2998 sec/batch
step: 3000/10000... loss: 3.6238... 0.3115 sec/batch
step: 3010/10000... loss: 3.7016... 0.3056 sec/batch
step: 3020/10000... loss: 3.6371... 0.3120 sec/batch
step: 3030/10000... loss: 3.6463... 0.3084 sec/batch
step: 3040/10000... loss: 3.5223... 0.3268 sec/batch
step: 3050/10000... loss: 3.6954... 0.3158 sec/batch
step: 3060/10000... loss: 3.7269... 0.3677 sec/batch
step: 3070/10000... loss: 3.7544... 0.3957 sec/batch
step: 3080/10000... loss: 3.6913... 0.3562 sec/batch
step: 3090/10000... loss: 3.6944... 0.3611 sec/batch
step: 3100/10000... loss: 3.6625... 0.3739 sec/batch
step: 3110/10000... loss: 3.6878... 0.3721 sec/batch
step: 3120/10000... loss: 3.7645... 0.3676 sec/batch
step: 3130/10000... loss: 3.5811... 0.3167 sec/batch
step: 3140/10000... loss: 3.6605... 0.3290 sec/batch
step: 3150/10000... loss: 3.6328... 0.3049 sec/batch
step: 3160/10000... loss: 3.7845... 0.3226 sec/batch
step: 3170/10000... loss: 3.6172... 0.3335 sec/batch
step: 3180/10000... loss: 3.6378... 0.3194 sec/batch
step: 3190/10000... loss: 3.7242... 0.3170 sec/batch
step: 3200/10000... loss: 3.5458... 0.3236 sec/batch
step: 3210/10000... loss: 3.5956... 0.3755 sec/batch
step: 3220/10000... loss: 3.8011... 0.4445 sec/batch
step: 3230/10000... loss: 3.7654... 0.3869 sec/batch
step: 3240/10000... loss: 3.5079... 0.3623 sec/batch
step: 3250/10000... loss: 3.5770... 0.3602 sec/batch
step: 3260/10000... loss: 3.5869... 0.4226 sec/batch
step: 3270/10000... loss: 3.6543... 0.3938 sec/batch
step: 3280/10000... loss: 3.6195... 0.3843 sec/batch
step: 3290/10000... loss: 3.6389... 0.4103 sec/batch
step: 3300/10000... loss: 3.4999... 0.3164 sec/batch
step: 3310/10000... loss: 3.6357... 0.3079 sec/batch
step: 3320/10000... loss: 3.7465... 0.3631 sec/batch
step: 3330/10000... loss: 3.6985... 0.3751 sec/batch
step: 3340/10000... loss: 3.6175... 0.3818 sec/batch
step: 3350/10000... loss: 3.6448... 0.3775 sec/batch
step: 3360/10000... loss: 3.6015... 0.3670 sec/batch
step: 3370/10000... loss: 3.6929... 0.4190 sec/batch
step: 3380/10000... loss: 3.7085... 0.3807 sec/batch
step: 3390/10000... loss: 3.5021... 0.4095 sec/batch
step: 3400/10000... loss: 3.6344... 0.4150 sec/batch
step: 3410/10000... loss: 3.5670... 0.3901 sec/batch
step: 3420/10000... loss: 3.7539... 0.4536 sec/batch
step: 3430/10000... loss: 3.6039... 0.3896 sec/batch
step: 3440/10000... loss: 3.6279... 0.3755 sec/batch
step: 3450/10000... loss: 3.6362... 0.3023 sec/batch
step: 3460/10000... loss: 3.4979... 0.3629 sec/batch
step: 3470/10000... loss: 3.5314... 0.4299 sec/batch
step: 3480/10000... loss: 3.7321... 0.3211 sec/batch
step: 3490/10000... loss: 3.7412... 0.3244 sec/batch
step: 3500/10000... loss: 3.4947... 0.3683 sec/batch
step: 3510/10000... loss: 3.5141... 0.3716 sec/batch
step: 3520/10000... loss: 3.5364... 0.3516 sec/batch
step: 3530/10000... loss: 3.6140... 0.3702 sec/batch
step: 3540/10000... loss: 3.5687... 0.3544 sec/batch
step: 3550/10000... loss: 3.6102... 0.3406 sec/batch
step: 3560/10000... loss: 3.4233... 0.3541 sec/batch
step: 3570/10000... loss: 3.5657... 0.3505 sec/batch
step: 3580/10000... loss: 3.6225... 0.3439 sec/batch
step: 3590/10000... loss: 3.6751... 0.3588 sec/batch
step: 3600/10000... loss: 3.6017... 0.3544 sec/batch
step: 3610/10000... loss: 3.5809... 0.3294 sec/batch
step: 3620/10000... loss: 3.5652... 0.3613 sec/batch
step: 3630/10000... loss: 3.6398... 0.3501 sec/batch
step: 3640/10000... loss: 3.6923... 0.3552 sec/batch
step: 3650/10000... loss: 3.4996... 0.3670 sec/batch
step: 3660/10000... loss: 3.6042... 0.3622 sec/batch
step: 3670/10000... loss: 3.5446... 0.3551 sec/batch
step: 3680/10000... loss: 3.7389... 0.3762 sec/batch
step: 3690/10000... loss: 3.5655... 0.3931 sec/batch
step: 3700/10000... loss: 3.5517... 0.3599 sec/batch
step: 3710/10000... loss: 3.6009... 0.4015 sec/batch
step: 3720/10000... loss: 3.4731... 0.3581 sec/batch
step: 3730/10000... loss: 3.4758... 0.3589 sec/batch
step: 3740/10000... loss: 3.7071... 0.3726 sec/batch
step: 3750/10000... loss: 3.6836... 0.3412 sec/batch
step: 3760/10000... loss: 3.4186... 0.3729 sec/batch
step: 3770/10000... loss: 3.4621... 0.3618 sec/batch
step: 3780/10000... loss: 3.5233... 0.3525 sec/batch
step: 3790/10000... loss: 3.5858... 0.4365 sec/batch
step: 3800/10000... loss: 3.5581... 0.3597 sec/batch
step: 3810/10000... loss: 3.5676... 0.3830 sec/batch
step: 3820/10000... loss: 3.3749... 0.3622 sec/batch
step: 3830/10000... loss: 3.5425... 0.3511 sec/batch
step: 3840/10000... loss: 3.6027... 0.3641 sec/batch
step: 3850/10000... loss: 3.5863... 0.3921 sec/batch
step: 3860/10000... loss: 3.5658... 0.3528 sec/batch
step: 3870/10000... loss: 3.5505... 0.3658 sec/batch
step: 3880/10000... loss: 3.5048... 0.3923 sec/batch
step: 3890/10000... loss: 3.6073... 0.3446 sec/batch
step: 3900/10000... loss: 3.6494... 0.3653 sec/batch
step: 3910/10000... loss: 3.4420... 0.4002 sec/batch
step: 3920/10000... loss: 3.5909... 0.4273 sec/batch
step: 3930/10000... loss: 3.4784... 0.3736 sec/batch
step: 3940/10000... loss: 3.6948... 0.3885 sec/batch
step: 3950/10000... loss: 3.5120... 0.4365 sec/batch
step: 3960/10000... loss: 3.5157... 0.3664 sec/batch
step: 3970/10000... loss: 3.5811... 0.3903 sec/batch
step: 3980/10000... loss: 3.4422... 0.3553 sec/batch
step: 3990/10000... loss: 3.4418... 0.3996 sec/batch
step: 4000/10000... loss: 3.6450... 0.3645 sec/batch
step: 4010/10000... loss: 3.6575... 0.3892 sec/batch
step: 4020/10000... loss: 3.4203... 0.3856 sec/batch
step: 4030/10000... loss: 3.4925... 0.3488 sec/batch
step: 4040/10000... loss: 3.4452... 0.3670 sec/batch
step: 4050/10000... loss: 3.5409... 0.4199 sec/batch
step: 4060/10000... loss: 3.5473... 0.3705 sec/batch
step: 4070/10000... loss: 3.5583... 0.4388 sec/batch
step: 4080/10000... loss: 3.3464... 0.3652 sec/batch
step: 4090/10000... loss: 3.5060... 0.3868 sec/batch
step: 4100/10000... loss: 3.5980... 0.4088 sec/batch
step: 4110/10000... loss: 3.6017... 0.3943 sec/batch
step: 4120/10000... loss: 3.5265... 0.4389 sec/batch
step: 4130/10000... loss: 3.5541... 0.3706 sec/batch
step: 4140/10000... loss: 3.4836... 0.4084 sec/batch
step: 4150/10000... loss: 3.5499... 0.3914 sec/batch
step: 4160/10000... loss: 3.6449... 0.3601 sec/batch
step: 4170/10000... loss: 3.4058... 0.3920 sec/batch
step: 4180/10000... loss: 3.5509... 0.3675 sec/batch
step: 4190/10000... loss: 3.4596... 0.3673 sec/batch
step: 4200/10000... loss: 3.6134... 0.3607 sec/batch
step: 4210/10000... loss: 3.4812... 0.3882 sec/batch
step: 4220/10000... loss: 3.4674... 0.3668 sec/batch
step: 4230/10000... loss: 3.5625... 0.4063 sec/batch
step: 4240/10000... loss: 3.4558... 0.3764 sec/batch
step: 4250/10000... loss: 3.4148... 0.3829 sec/batch
step: 4260/10000... loss: 3.6311... 0.4135 sec/batch
step: 4270/10000... loss: 3.6153... 0.3798 sec/batch
step: 4280/10000... loss: 3.3791... 0.3695 sec/batch
step: 4290/10000... loss: 3.4043... 0.3923 sec/batch
step: 4300/10000... loss: 3.4486... 0.3474 sec/batch
step: 4310/10000... loss: 3.5095... 0.3989 sec/batch
step: 4320/10000... loss: 3.4933... 0.3395 sec/batch
step: 4330/10000... loss: 3.4983... 0.3052 sec/batch
step: 4340/10000... loss: 3.3339... 0.3242 sec/batch
step: 4350/10000... loss: 3.4939... 0.3662 sec/batch
step: 4360/10000... loss: 3.5618... 0.3782 sec/batch
step: 4370/10000... loss: 3.5442... 0.3500 sec/batch
step: 4380/10000... loss: 3.4714... 0.3583 sec/batch
step: 4390/10000... loss: 3.4945... 0.3964 sec/batch
step: 4400/10000... loss: 3.5040... 0.3774 sec/batch
step: 4410/10000... loss: 3.5228... 0.3872 sec/batch
step: 4420/10000... loss: 3.6105... 0.3688 sec/batch
step: 4430/10000... loss: 3.4255... 0.3729 sec/batch
step: 4440/10000... loss: 3.5174... 0.4281 sec/batch
step: 4450/10000... loss: 3.4510... 0.3880 sec/batch
step: 4460/10000... loss: 3.6152... 0.3897 sec/batch
step: 4470/10000... loss: 3.4354... 0.3757 sec/batch
step: 4480/10000... loss: 3.4908... 0.3257 sec/batch
step: 4490/10000... loss: 3.4943... 0.3138 sec/batch
step: 4500/10000... loss: 3.3717... 0.3198 sec/batch
step: 4510/10000... loss: 3.3864... 0.2946 sec/batch
step: 4520/10000... loss: 3.5899... 0.3189 sec/batch
step: 4530/10000... loss: 3.6222... 0.3917 sec/batch
step: 4540/10000... loss: 3.3392... 0.3573 sec/batch
step: 4550/10000... loss: 3.4117... 0.3127 sec/batch
step: 4560/10000... loss: 3.4366... 0.3660 sec/batch
step: 4570/10000... loss: 3.4827... 0.3556 sec/batch
step: 4580/10000... loss: 3.4514... 0.3864 sec/batch
step: 4590/10000... loss: 3.4926... 0.3834 sec/batch
step: 4600/10000... loss: 3.3017... 0.3839 sec/batch
step: 4610/10000... loss: 3.4654... 0.3662 sec/batch
step: 4620/10000... loss: 3.5270... 0.3977 sec/batch
step: 4630/10000... loss: 3.5356... 0.3466 sec/batch
step: 4640/10000... loss: 3.4969... 0.4300 sec/batch
step: 4650/10000... loss: 3.4679... 0.2964 sec/batch
step: 4660/10000... loss: 3.4611... 0.3296 sec/batch
step: 4670/10000... loss: 3.4977... 0.3145 sec/batch
step: 4680/10000... loss: 3.5319... 0.3183 sec/batch
step: 4690/10000... loss: 3.3274... 0.3256 sec/batch
step: 4700/10000... loss: 3.4967... 0.3439 sec/batch
step: 4710/10000... loss: 3.4207... 0.3054 sec/batch
step: 4720/10000... loss: 3.6008... 0.3277 sec/batch
step: 4730/10000... loss: 3.4199... 0.3712 sec/batch
step: 4740/10000... loss: 3.4551... 0.3763 sec/batch
step: 4750/10000... loss: 3.5120... 0.3972 sec/batch
step: 4760/10000... loss: 3.3721... 0.3955 sec/batch
step: 4770/10000... loss: 3.4041... 0.4017 sec/batch
step: 4780/10000... loss: 3.5847... 0.3850 sec/batch
step: 4790/10000... loss: 3.5421... 0.3794 sec/batch
step: 4800/10000... loss: 3.3377... 0.3737 sec/batch
step: 4810/10000... loss: 3.3812... 0.3782 sec/batch
step: 4820/10000... loss: 3.4376... 0.3791 sec/batch
step: 4830/10000... loss: 3.5027... 0.3189 sec/batch
step: 4840/10000... loss: 3.4538... 0.3023 sec/batch
step: 4850/10000... loss: 3.4184... 0.3348 sec/batch
step: 4860/10000... loss: 3.2584... 0.3110 sec/batch
step: 4870/10000... loss: 3.4880... 0.3195 sec/batch
step: 4880/10000... loss: 3.5125... 0.3105 sec/batch
step: 4890/10000... loss: 3.5457... 0.3758 sec/batch
step: 4900/10000... loss: 3.4830... 0.4058 sec/batch
step: 4910/10000... loss: 3.4547... 0.3711 sec/batch
step: 4920/10000... loss: 3.4324... 0.3788 sec/batch
step: 4930/10000... loss: 3.4775... 0.3927 sec/batch
step: 4940/10000... loss: 3.5052... 0.3912 sec/batch
step: 4950/10000... loss: 3.3535... 0.3704 sec/batch
step: 4960/10000... loss: 3.4230... 0.3875 sec/batch
step: 4970/10000... loss: 3.3973... 0.4310 sec/batch
step: 4980/10000... loss: 3.5837... 0.3125 sec/batch
step: 4990/10000... loss: 3.4003... 0.3081 sec/batch
step: 5000/10000... loss: 3.4108... 0.4092 sec/batch
step: 5010/10000... loss: 3.4400... 0.3835 sec/batch
step: 5020/10000... loss: 3.3382... 0.3487 sec/batch
step: 5030/10000... loss: 3.3305... 0.3862 sec/batch
step: 5040/10000... loss: 3.5593... 0.3383 sec/batch
step: 5050/10000... loss: 3.5198... 0.3558 sec/batch
step: 5060/10000... loss: 3.3355... 0.3444 sec/batch
step: 5070/10000... loss: 3.3671... 0.3571 sec/batch
step: 5080/10000... loss: 3.3856... 0.3578 sec/batch
step: 5090/10000... loss: 3.4277... 0.4110 sec/batch
step: 5100/10000... loss: 3.4212... 0.4329 sec/batch
step: 5110/10000... loss: 3.3859... 0.3715 sec/batch
step: 5120/10000... loss: 3.2650... 0.3706 sec/batch
step: 5130/10000... loss: 3.4748... 0.3229 sec/batch
step: 5140/10000... loss: 3.4633... 0.3215 sec/batch
step: 5150/10000... loss: 3.5263... 0.3067 sec/batch
step: 5160/10000... loss: 3.4294... 0.3112 sec/batch
step: 5170/10000... loss: 3.3648... 0.3053 sec/batch
step: 5180/10000... loss: 3.4044... 0.3225 sec/batch
step: 5190/10000... loss: 3.4552... 0.3101 sec/batch
step: 5200/10000... loss: 3.4801... 0.3283 sec/batch
step: 5210/10000... loss: 3.2786... 0.3182 sec/batch
step: 5220/10000... loss: 3.4250... 0.3253 sec/batch
step: 5230/10000... loss: 3.3599... 0.3119 sec/batch
step: 5240/10000... loss: 3.5392... 0.3324 sec/batch
step: 5250/10000... loss: 3.4000... 0.3680 sec/batch
step: 5260/10000... loss: 3.4063... 0.3915 sec/batch
step: 5270/10000... loss: 3.4498... 0.3764 sec/batch
step: 5280/10000... loss: 3.3663... 0.3701 sec/batch
step: 5290/10000... loss: 3.3090... 0.4116 sec/batch
step: 5300/10000... loss: 3.5301... 0.4324 sec/batch
step: 5310/10000... loss: 3.4767... 0.3846 sec/batch
step: 5320/10000... loss: 3.3157... 0.4349 sec/batch
step: 5330/10000... loss: 3.3302... 0.3677 sec/batch
step: 5340/10000... loss: 3.3771... 0.3572 sec/batch
step: 5350/10000... loss: 3.4148... 0.4051 sec/batch
step: 5360/10000... loss: 3.4040... 0.3563 sec/batch
step: 5370/10000... loss: 3.3973... 0.3951 sec/batch
step: 5380/10000... loss: 3.2471... 0.3549 sec/batch
step: 5390/10000... loss: 3.4461... 0.3459 sec/batch
step: 5400/10000... loss: 3.4701... 0.3427 sec/batch
step: 5410/10000... loss: 3.4626... 0.3663 sec/batch
step: 5420/10000... loss: 3.4472... 0.3126 sec/batch
step: 5430/10000... loss: 3.3845... 0.3856 sec/batch
step: 5440/10000... loss: 3.3408... 0.4253 sec/batch
step: 5450/10000... loss: 3.4366... 0.4196 sec/batch
step: 5460/10000... loss: 3.4623... 0.4029 sec/batch
step: 5470/10000... loss: 3.2996... 0.3884 sec/batch
step: 5480/10000... loss: 3.3516... 0.3840 sec/batch
step: 5490/10000... loss: 3.3840... 0.3780 sec/batch
step: 5500/10000... loss: 3.4830... 0.3930 sec/batch
step: 5510/10000... loss: 3.3992... 0.3928 sec/batch
step: 5520/10000... loss: 3.3723... 0.3844 sec/batch
step: 5530/10000... loss: 3.4530... 0.3582 sec/batch
step: 5540/10000... loss: 3.2945... 0.3500 sec/batch
step: 5550/10000... loss: 3.2469... 0.4040 sec/batch
step: 5560/10000... loss: 3.4846... 0.3912 sec/batch
step: 5570/10000... loss: 3.4699... 0.3838 sec/batch
step: 5580/10000... loss: 3.2403... 0.4273 sec/batch
step: 5590/10000... loss: 3.2730... 0.4352 sec/batch
step: 5600/10000... loss: 3.3511... 0.3581 sec/batch
step: 5610/10000... loss: 3.3692... 0.3985 sec/batch
step: 5620/10000... loss: 3.3898... 0.3772 sec/batch
step: 5630/10000... loss: 3.4274... 0.3458 sec/batch
step: 5640/10000... loss: 3.2729... 0.3627 sec/batch
step: 5650/10000... loss: 3.4467... 0.4096 sec/batch
step: 5660/10000... loss: 3.4666... 0.4255 sec/batch
step: 5670/10000... loss: 3.4247... 0.3962 sec/batch
step: 5680/10000... loss: 3.4026... 0.3790 sec/batch
step: 5690/10000... loss: 3.3936... 0.3995 sec/batch
step: 5700/10000... loss: 3.3496... 0.3813 sec/batch
step: 5710/10000... loss: 3.4115... 0.3683 sec/batch
step: 5720/10000... loss: 3.4696... 0.4006 sec/batch
step: 5730/10000... loss: 3.2536... 0.3981 sec/batch
step: 5740/10000... loss: 3.3771... 0.3488 sec/batch
step: 5750/10000... loss: 3.3523... 0.3918 sec/batch
step: 5760/10000... loss: 3.4490... 0.3487 sec/batch
step: 5770/10000... loss: 3.3559... 0.4229 sec/batch
step: 5780/10000... loss: 3.3325... 0.3629 sec/batch
step: 5790/10000... loss: 3.3738... 0.4115 sec/batch
step: 5800/10000... loss: 3.2481... 0.4140 sec/batch
step: 5810/10000... loss: 3.2923... 0.3702 sec/batch
step: 5820/10000... loss: 3.4949... 0.3566 sec/batch
step: 5830/10000... loss: 3.4618... 0.3504 sec/batch
step: 5840/10000... loss: 3.2457... 0.4090 sec/batch
step: 5850/10000... loss: 3.2862... 0.3475 sec/batch
step: 5860/10000... loss: 3.3000... 0.3543 sec/batch
step: 5870/10000... loss: 3.3617... 0.3503 sec/batch
step: 5880/10000... loss: 3.3439... 0.3924 sec/batch
step: 5890/10000... loss: 3.3554... 0.3460 sec/batch
step: 5900/10000... loss: 3.2021... 0.3680 sec/batch
step: 5910/10000... loss: 3.3534... 0.3841 sec/batch
step: 5920/10000... loss: 3.4368... 0.3605 sec/batch
step: 5930/10000... loss: 3.4410... 0.3740 sec/batch
step: 5940/10000... loss: 3.3300... 0.3518 sec/batch
step: 5950/10000... loss: 3.3520... 0.3815 sec/batch
step: 5960/10000... loss: 3.3190... 0.3546 sec/batch
step: 5970/10000... loss: 3.3669... 0.3721 sec/batch
step: 5980/10000... loss: 3.4609... 0.3432 sec/batch
step: 5990/10000... loss: 3.2475... 0.3620 sec/batch
step: 6000/10000... loss: 3.3769... 0.3821 sec/batch
step: 6010/10000... loss: 3.3321... 0.3659 sec/batch
step: 6020/10000... loss: 3.4747... 0.4085 sec/batch
step: 6030/10000... loss: 3.3160... 0.3791 sec/batch
step: 6040/10000... loss: 3.3289... 0.3743 sec/batch
step: 6050/10000... loss: 3.3784... 0.4019 sec/batch
step: 6060/10000... loss: 3.2490... 0.3849 sec/batch
step: 6070/10000... loss: 3.2661... 0.3927 sec/batch
step: 6080/10000... loss: 3.4885... 0.3965 sec/batch
step: 6090/10000... loss: 3.4254... 0.3717 sec/batch
step: 6100/10000... loss: 3.2344... 0.3030 sec/batch
step: 6110/10000... loss: 3.3007... 0.3348 sec/batch
step: 6120/10000... loss: 3.2880... 0.3160 sec/batch
step: 6130/10000... loss: 3.3619... 0.3381 sec/batch
step: 6140/10000... loss: 3.3051... 0.3053 sec/batch
step: 6150/10000... loss: 3.3559... 0.3236 sec/batch
step: 6160/10000... loss: 3.2439... 0.3215 sec/batch
step: 6170/10000... loss: 3.3956... 0.3138 sec/batch
step: 6180/10000... loss: 3.4223... 0.3147 sec/batch
step: 6190/10000... loss: 3.3937... 0.3109 sec/batch
step: 6200/10000... loss: 3.3077... 0.3306 sec/batch
step: 6210/10000... loss: 3.3813... 0.3105 sec/batch
step: 6220/10000... loss: 3.3416... 0.3291 sec/batch
step: 6230/10000... loss: 3.3803... 0.3147 sec/batch
step: 6240/10000... loss: 3.4179... 0.3120 sec/batch
step: 6250/10000... loss: 3.2255... 0.3067 sec/batch
step: 6260/10000... loss: 3.3499... 0.3209 sec/batch
step: 6270/10000... loss: 3.3163... 0.3269 sec/batch
step: 6280/10000... loss: 3.4701... 0.3271 sec/batch
step: 6290/10000... loss: 3.3207... 0.3111 sec/batch
step: 6300/10000... loss: 3.3710... 0.3133 sec/batch
step: 6310/10000... loss: 3.3750... 0.3347 sec/batch
step: 6320/10000... loss: 3.2542... 0.3344 sec/batch
step: 6330/10000... loss: 3.2631... 0.3064 sec/batch
step: 6340/10000... loss: 3.4645... 0.3112 sec/batch
step: 6350/10000... loss: 3.4444... 0.3231 sec/batch
step: 6360/10000... loss: 3.2248... 0.3036 sec/batch
step: 6370/10000... loss: 3.2514... 0.3341 sec/batch
step: 6380/10000... loss: 3.2590... 0.3176 sec/batch
step: 6390/10000... loss: 3.3279... 0.3506 sec/batch
step: 6400/10000... loss: 3.2932... 0.3108 sec/batch
step: 6410/10000... loss: 3.3311... 0.3137 sec/batch
step: 6420/10000... loss: 3.1643... 0.3289 sec/batch
step: 6430/10000... loss: 3.3686... 0.3065 sec/batch
step: 6440/10000... loss: 3.3933... 0.3287 sec/batch
step: 6450/10000... loss: 3.4045... 0.3112 sec/batch
step: 6460/10000... loss: 3.3084... 0.3138 sec/batch
step: 6470/10000... loss: 3.3150... 0.3139 sec/batch
step: 6480/10000... loss: 3.3649... 0.3088 sec/batch
step: 6490/10000... loss: 3.3379... 0.3523 sec/batch
step: 6500/10000... loss: 3.4286... 0.3279 sec/batch
step: 6510/10000... loss: 3.2447... 0.3506 sec/batch
step: 6520/10000... loss: 3.3421... 0.3250 sec/batch
step: 6530/10000... loss: 3.3450... 0.3053 sec/batch
step: 6540/10000... loss: 3.4325... 0.3180 sec/batch
step: 6550/10000... loss: 3.3024... 0.3528 sec/batch
step: 6560/10000... loss: 3.3715... 0.3284 sec/batch
step: 6570/10000... loss: 3.3495... 0.3932 sec/batch
step: 6580/10000... loss: 3.2122... 0.3251 sec/batch
step: 6590/10000... loss: 3.2157... 0.3207 sec/batch
step: 6600/10000... loss: 3.4787... 0.3354 sec/batch
step: 6610/10000... loss: 3.4142... 0.3439 sec/batch
step: 6620/10000... loss: 3.1633... 0.3128 sec/batch
step: 6630/10000... loss: 3.2525... 0.3254 sec/batch
step: 6640/10000... loss: 3.2499... 0.3260 sec/batch
step: 6650/10000... loss: 3.3293... 0.3192 sec/batch
step: 6660/10000... loss: 3.3072... 0.3318 sec/batch
step: 6670/10000... loss: 3.2907... 0.3594 sec/batch
step: 6680/10000... loss: 3.1923... 0.3311 sec/batch
step: 6690/10000... loss: 3.3300... 0.3062 sec/batch
step: 6700/10000... loss: 3.4043... 0.3304 sec/batch
step: 6710/10000... loss: 3.3489... 0.3185 sec/batch
step: 6720/10000... loss: 3.3200... 0.3196 sec/batch
step: 6730/10000... loss: 3.3080... 0.3190 sec/batch
step: 6740/10000... loss: 3.3215... 0.3059 sec/batch
step: 6750/10000... loss: 3.3977... 0.3515 sec/batch
step: 6760/10000... loss: 3.3721... 0.3285 sec/batch
step: 6770/10000... loss: 3.1774... 0.3173 sec/batch
step: 6780/10000... loss: 3.3492... 0.3293 sec/batch
step: 6790/10000... loss: 3.3130... 0.3115 sec/batch
step: 6800/10000... loss: 3.4374... 0.3130 sec/batch
step: 6810/10000... loss: 3.2965... 0.3416 sec/batch
step: 6820/10000... loss: 3.3148... 0.4105 sec/batch
step: 6830/10000... loss: 3.3264... 0.3053 sec/batch
step: 6840/10000... loss: 3.2276... 0.3404 sec/batch
step: 6850/10000... loss: 3.2099... 0.3183 sec/batch
step: 6860/10000... loss: 3.4336... 0.3408 sec/batch
step: 6870/10000... loss: 3.4153... 0.3371 sec/batch
step: 6880/10000... loss: 3.1623... 0.3056 sec/batch
step: 6890/10000... loss: 3.2241... 0.3350 sec/batch
step: 6900/10000... loss: 3.2443... 0.3234 sec/batch
step: 6910/10000... loss: 3.3349... 0.3216 sec/batch
step: 6920/10000... loss: 3.2614... 0.3445 sec/batch
step: 6930/10000... loss: 3.2930... 0.3090 sec/batch
step: 6940/10000... loss: 3.1615... 0.3176 sec/batch
step: 6950/10000... loss: 3.3637... 0.3174 sec/batch
step: 6960/10000... loss: 3.3576... 0.3260 sec/batch
step: 6970/10000... loss: 3.3851... 0.3234 sec/batch
step: 6980/10000... loss: 3.3164... 0.3032 sec/batch
step: 6990/10000... loss: 3.3072... 0.3066 sec/batch
step: 7000/10000... loss: 3.2948... 0.3185 sec/batch
step: 7010/10000... loss: 3.3539... 0.4196 sec/batch
step: 7020/10000... loss: 3.4111... 0.3143 sec/batch
step: 7030/10000... loss: 3.1734... 0.3238 sec/batch
step: 7040/10000... loss: 3.3039... 0.3120 sec/batch
step: 7050/10000... loss: 3.2881... 0.3145 sec/batch
step: 7060/10000... loss: 3.4290... 0.3494 sec/batch
step: 7070/10000... loss: 3.2524... 0.2973 sec/batch
step: 7080/10000... loss: 3.2703... 0.3203 sec/batch
step: 7090/10000... loss: 3.3012... 0.3318 sec/batch
step: 7100/10000... loss: 3.2307... 0.3107 sec/batch
step: 7110/10000... loss: 3.1877... 0.3396 sec/batch
step: 7120/10000... loss: 3.4140... 0.3262 sec/batch
step: 7130/10000... loss: 3.3763... 0.3139 sec/batch
step: 7140/10000... loss: 3.1230... 0.3152 sec/batch
step: 7150/10000... loss: 3.2333... 0.3522 sec/batch
step: 7160/10000... loss: 3.2506... 0.3646 sec/batch
step: 7170/10000... loss: 3.3220... 0.3139 sec/batch
step: 7180/10000... loss: 3.2422... 0.3068 sec/batch
step: 7190/10000... loss: 3.2480... 0.3207 sec/batch
step: 7200/10000... loss: 3.1786... 0.3233 sec/batch
step: 7210/10000... loss: 3.3310... 0.3121 sec/batch
step: 7220/10000... loss: 3.3874... 0.3278 sec/batch
step: 7230/10000... loss: 3.3027... 0.3179 sec/batch
step: 7240/10000... loss: 3.3132... 0.3427 sec/batch
step: 7250/10000... loss: 3.2664... 0.3417 sec/batch
step: 7260/10000... loss: 3.2775... 0.3235 sec/batch
step: 7270/10000... loss: 3.3208... 0.3332 sec/batch
step: 7280/10000... loss: 3.4209... 0.3155 sec/batch
step: 7290/10000... loss: 3.0895... 0.3215 sec/batch
step: 7300/10000... loss: 3.3185... 0.3534 sec/batch
step: 7310/10000... loss: 3.2558... 0.3065 sec/batch
step: 7320/10000... loss: 3.4093... 0.3238 sec/batch
step: 7330/10000... loss: 3.2516... 0.3306 sec/batch
step: 7340/10000... loss: 3.2992... 0.3298 sec/batch
step: 7350/10000... loss: 3.3422... 0.3072 sec/batch
step: 7360/10000... loss: 3.2042... 0.3171 sec/batch
step: 7370/10000... loss: 3.1816... 0.3154 sec/batch
step: 7380/10000... loss: 3.3904... 0.3123 sec/batch
step: 7390/10000... loss: 3.3974... 0.3200 sec/batch
step: 7400/10000... loss: 3.1356... 0.3213 sec/batch
step: 7410/10000... loss: 3.2551... 0.3380 sec/batch
step: 7420/10000... loss: 3.2382... 0.3283 sec/batch
step: 7430/10000... loss: 3.2855... 0.3237 sec/batch
step: 7440/10000... loss: 3.2191... 0.3399 sec/batch
step: 7450/10000... loss: 3.2972... 0.3390 sec/batch
step: 7460/10000... loss: 3.1575... 0.3200 sec/batch
step: 7470/10000... loss: 3.2988... 0.3181 sec/batch
step: 7480/10000... loss: 3.3620... 0.3208 sec/batch
step: 7490/10000... loss: 3.3343... 0.3128 sec/batch
step: 7500/10000... loss: 3.2574... 0.3214 sec/batch
step: 7510/10000... loss: 3.2677... 0.3276 sec/batch
step: 7520/10000... loss: 3.2838... 0.3097 sec/batch
step: 7530/10000... loss: 3.3416... 0.3070 sec/batch
step: 7540/10000... loss: 3.3776... 0.3113 sec/batch
step: 7550/10000... loss: 3.1506... 0.3052 sec/batch
step: 7560/10000... loss: 3.2498... 0.3318 sec/batch
step: 7570/10000... loss: 3.2316... 0.3221 sec/batch
step: 7580/10000... loss: 3.3557... 0.3145 sec/batch
step: 7590/10000... loss: 3.2678... 0.3033 sec/batch
step: 7600/10000... loss: 3.2692... 0.3143 sec/batch
step: 7610/10000... loss: 3.2650... 0.3183 sec/batch
step: 7620/10000... loss: 3.2266... 0.3047 sec/batch
step: 7630/10000... loss: 3.1280... 0.3354 sec/batch
step: 7640/10000... loss: 3.3648... 0.3009 sec/batch
step: 7650/10000... loss: 3.3885... 0.3375 sec/batch
step: 7660/10000... loss: 3.1392... 0.3232 sec/batch
step: 7670/10000... loss: 3.1840... 0.3070 sec/batch
step: 7680/10000... loss: 3.1947... 0.3134 sec/batch
step: 7690/10000... loss: 3.2949... 0.3236 sec/batch
step: 7700/10000... loss: 3.2633... 0.3716 sec/batch
step: 7710/10000... loss: 3.2874... 0.3330 sec/batch
step: 7720/10000... loss: 3.1269... 0.3123 sec/batch
step: 7730/10000... loss: 3.2974... 0.3159 sec/batch
step: 7740/10000... loss: 3.3255... 0.3126 sec/batch
step: 7750/10000... loss: 3.3594... 0.3117 sec/batch
step: 7760/10000... loss: 3.2584... 0.3169 sec/batch
step: 7770/10000... loss: 3.2482... 0.3210 sec/batch
step: 7780/10000... loss: 3.2560... 0.3386 sec/batch
step: 7790/10000... loss: 3.3025... 0.3127 sec/batch
step: 7800/10000... loss: 3.3511... 0.3091 sec/batch
step: 7810/10000... loss: 3.1368... 0.3243 sec/batch
step: 7820/10000... loss: 3.2819... 0.3078 sec/batch
step: 7830/10000... loss: 3.2117... 0.3363 sec/batch
step: 7840/10000... loss: 3.3885... 0.3629 sec/batch
step: 7850/10000... loss: 3.1925... 0.3302 sec/batch
step: 7860/10000... loss: 3.2918... 0.3187 sec/batch
step: 7870/10000... loss: 3.2930... 0.3029 sec/batch
step: 7880/10000... loss: 3.1710... 0.3149 sec/batch
step: 7890/10000... loss: 3.1881... 0.3157 sec/batch
step: 7900/10000... loss: 3.3762... 0.3466 sec/batch
step: 7910/10000... loss: 3.3545... 0.3503 sec/batch
step: 7920/10000... loss: 3.1478... 0.3505 sec/batch
step: 7930/10000... loss: 3.2087... 0.3150 sec/batch
step: 7940/10000... loss: 3.1928... 0.3506 sec/batch
step: 7950/10000... loss: 3.3092... 0.3103 sec/batch
step: 7960/10000... loss: 3.2053... 0.3158 sec/batch
step: 7970/10000... loss: 3.2184... 0.3281 sec/batch
step: 7980/10000... loss: 3.1255... 0.3679 sec/batch
step: 7990/10000... loss: 3.2897... 0.3040 sec/batch
step: 8000/10000... loss: 3.3263... 0.3155 sec/batch
step: 8010/10000... loss: 3.3062... 0.3593 sec/batch
step: 8020/10000... loss: 3.2555... 0.3240 sec/batch
step: 8030/10000... loss: 3.2169... 0.3037 sec/batch
step: 8040/10000... loss: 3.2185... 0.3152 sec/batch
step: 8050/10000... loss: 3.3121... 0.3085 sec/batch
step: 8060/10000... loss: 3.3904... 0.3239 sec/batch
step: 8070/10000... loss: 3.1385... 0.3365 sec/batch
step: 8080/10000... loss: 3.3019... 0.3539 sec/batch
step: 8090/10000... loss: 3.2182... 0.3623 sec/batch
step: 8100/10000... loss: 3.3644... 0.3144 sec/batch
step: 8110/10000... loss: 3.2205... 0.3284 sec/batch
step: 8120/10000... loss: 3.2473... 0.3095 sec/batch
step: 8130/10000... loss: 3.3031... 0.3243 sec/batch
step: 8140/10000... loss: 3.1995... 0.3195 sec/batch
step: 8150/10000... loss: 3.1387... 0.3412 sec/batch
step: 8160/10000... loss: 3.3365... 0.3195 sec/batch
step: 8170/10000... loss: 3.3399... 0.3286 sec/batch
step: 8180/10000... loss: 3.1500... 0.3577 sec/batch
step: 8190/10000... loss: 3.1968... 0.3006 sec/batch
step: 8200/10000... loss: 3.1907... 0.3115 sec/batch
step: 8210/10000... loss: 3.2628... 0.3181 sec/batch
step: 8220/10000... loss: 3.2180... 0.3045 sec/batch
step: 8230/10000... loss: 3.2431... 0.3318 sec/batch
step: 8240/10000... loss: 3.0974... 0.3462 sec/batch
step: 8250/10000... loss: 3.2945... 0.3203 sec/batch
step: 8260/10000... loss: 3.2867... 0.3567 sec/batch
step: 8270/10000... loss: 3.2817... 0.3076 sec/batch
step: 8280/10000... loss: 3.2499... 0.3079 sec/batch
step: 8290/10000... loss: 3.2344... 0.3708 sec/batch
step: 8300/10000... loss: 3.2364... 0.3089 sec/batch
step: 8310/10000... loss: 3.2509... 0.3232 sec/batch
step: 8320/10000... loss: 3.3110... 0.3160 sec/batch
step: 8330/10000... loss: 3.1442... 0.3091 sec/batch
step: 8340/10000... loss: 3.2928... 0.3310 sec/batch
step: 8350/10000... loss: 3.1794... 0.3647 sec/batch
step: 8360/10000... loss: 3.3666... 0.4161 sec/batch
step: 8370/10000... loss: 3.2324... 0.3220 sec/batch
step: 8380/10000... loss: 3.2170... 0.3330 sec/batch
step: 8390/10000... loss: 3.2769... 0.3128 sec/batch
step: 8400/10000... loss: 3.1591... 0.3092 sec/batch
step: 8410/10000... loss: 3.1293... 0.3091 sec/batch
step: 8420/10000... loss: 3.3152... 0.3335 sec/batch
step: 8430/10000... loss: 3.3480... 0.3927 sec/batch
step: 8440/10000... loss: 3.0834... 0.4002 sec/batch
step: 8450/10000... loss: 3.2049... 0.3251 sec/batch
step: 8460/10000... loss: 3.1784... 0.3294 sec/batch
step: 8470/10000... loss: 3.2458... 0.3164 sec/batch
step: 8480/10000... loss: 3.2333... 0.3315 sec/batch
step: 8490/10000... loss: 3.2118... 0.3200 sec/batch
step: 8500/10000... loss: 3.0838... 0.3049 sec/batch
step: 8510/10000... loss: 3.2669... 0.3615 sec/batch
step: 8520/10000... loss: 3.2896... 0.3804 sec/batch
step: 8530/10000... loss: 3.2551... 0.3274 sec/batch
step: 8540/10000... loss: 3.2157... 0.3148 sec/batch
step: 8550/10000... loss: 3.2087... 0.3640 sec/batch
step: 8560/10000... loss: 3.2282... 0.3081 sec/batch
step: 8570/10000... loss: 3.2541... 0.3384 sec/batch
step: 8580/10000... loss: 3.2606... 0.4182 sec/batch
step: 8590/10000... loss: 3.1056... 0.3084 sec/batch
step: 8600/10000... loss: 3.2377... 0.4165 sec/batch
step: 8610/10000... loss: 3.2136... 0.3227 sec/batch
step: 8620/10000... loss: 3.3749... 0.3076 sec/batch
step: 8630/10000... loss: 3.1993... 0.3176 sec/batch
step: 8640/10000... loss: 3.2638... 0.3352 sec/batch
step: 8650/10000... loss: 3.2294... 0.3578 sec/batch
step: 8660/10000... loss: 3.1542... 0.3055 sec/batch
step: 8670/10000... loss: 3.1177... 0.3082 sec/batch
step: 8680/10000... loss: 3.3315... 0.3306 sec/batch
step: 8690/10000... loss: 3.3303... 0.3340 sec/batch
step: 8700/10000... loss: 3.0779... 0.3048 sec/batch
step: 8710/10000... loss: 3.1721... 0.3125 sec/batch
step: 8720/10000... loss: 3.1968... 0.3133 sec/batch
step: 8730/10000... loss: 3.2540... 0.3075 sec/batch
step: 8740/10000... loss: 3.2391... 0.3262 sec/batch
step: 8750/10000... loss: 3.2135... 0.3191 sec/batch
step: 8760/10000... loss: 3.0627... 0.3260 sec/batch
step: 8770/10000... loss: 3.2197... 0.3059 sec/batch
step: 8780/10000... loss: 3.2424... 0.3762 sec/batch
step: 8790/10000... loss: 3.2671... 0.3259 sec/batch
step: 8800/10000... loss: 3.2056... 0.3234 sec/batch
step: 8810/10000... loss: 3.2270... 0.3070 sec/batch
step: 8820/10000... loss: 3.2309... 0.3220 sec/batch
step: 8830/10000... loss: 3.2549... 0.3322 sec/batch
step: 8840/10000... loss: 3.2991... 0.3184 sec/batch
step: 8850/10000... loss: 3.1412... 0.3274 sec/batch
step: 8860/10000... loss: 3.2178... 0.3722 sec/batch
step: 8870/10000... loss: 3.1530... 0.3162 sec/batch
step: 8880/10000... loss: 3.3398... 0.3231 sec/batch
step: 8890/10000... loss: 3.2101... 0.3702 sec/batch
step: 8900/10000... loss: 3.2393... 0.4616 sec/batch
step: 8910/10000... loss: 3.2250... 0.3629 sec/batch
step: 8920/10000... loss: 3.1390... 0.3224 sec/batch
step: 8930/10000... loss: 3.0451... 0.4069 sec/batch
step: 8940/10000... loss: 3.2857... 0.3391 sec/batch
step: 8950/10000... loss: 3.3494... 0.3126 sec/batch
step: 8960/10000... loss: 3.0726... 0.3015 sec/batch
step: 8970/10000... loss: 3.1308... 0.3245 sec/batch
step: 8980/10000... loss: 3.1821... 0.3063 sec/batch
step: 8990/10000... loss: 3.2460... 0.4287 sec/batch
step: 9000/10000... loss: 3.2032... 0.3020 sec/batch
step: 9010/10000... loss: 3.1954... 0.3316 sec/batch
step: 9020/10000... loss: 3.0546... 0.3135 sec/batch
step: 9030/10000... loss: 3.2273... 0.3923 sec/batch
step: 9040/10000... loss: 3.2736... 0.3033 sec/batch
step: 9050/10000... loss: 3.2285... 0.3443 sec/batch
step: 9060/10000... loss: 3.1928... 0.3186 sec/batch
step: 9070/10000... loss: 3.2211... 0.3363 sec/batch
step: 9080/10000... loss: 3.1986... 0.3015 sec/batch
step: 9090/10000... loss: 3.2688... 0.3588 sec/batch
step: 9100/10000... loss: 3.2571... 0.3823 sec/batch
step: 9110/10000... loss: 3.1084... 0.3762 sec/batch
step: 9120/10000... loss: 3.2348... 0.3612 sec/batch
step: 9130/10000... loss: 3.1911... 0.3721 sec/batch
step: 9140/10000... loss: 3.3045... 0.3695 sec/batch
step: 9150/10000... loss: 3.1692... 0.4018 sec/batch
step: 9160/10000... loss: 3.1913... 0.3192 sec/batch
step: 9170/10000... loss: 3.1722... 0.3437 sec/batch
step: 9180/10000... loss: 3.1678... 0.3619 sec/batch
step: 9190/10000... loss: 3.0893... 0.3371 sec/batch
step: 9200/10000... loss: 3.3129... 0.3595 sec/batch
step: 9210/10000... loss: 3.2847... 0.4454 sec/batch
step: 9220/10000... loss: 3.1234... 0.3550 sec/batch
step: 9230/10000... loss: 3.1202... 0.3632 sec/batch
step: 9240/10000... loss: 3.1642... 0.3493 sec/batch
step: 9250/10000... loss: 3.2155... 0.3782 sec/batch
step: 9260/10000... loss: 3.1908... 0.3097 sec/batch
step: 9270/10000... loss: 3.2232... 0.3924 sec/batch
step: 9280/10000... loss: 3.0476... 0.3672 sec/batch
step: 9290/10000... loss: 3.1980... 0.3843 sec/batch
step: 9300/10000... loss: 3.2316... 0.3299 sec/batch
step: 9310/10000... loss: 3.2401... 0.3079 sec/batch
step: 9320/10000... loss: 3.2051... 0.3077 sec/batch
step: 9330/10000... loss: 3.1876... 0.3002 sec/batch
step: 9340/10000... loss: 3.2034... 0.3062 sec/batch
step: 9350/10000... loss: 3.2517... 0.3177 sec/batch
step: 9360/10000... loss: 3.2540... 0.3104 sec/batch
step: 9370/10000... loss: 3.0868... 0.3347 sec/batch
step: 9380/10000... loss: 3.1699... 0.3104 sec/batch
step: 9390/10000... loss: 3.1867... 0.3267 sec/batch
step: 9400/10000... loss: 3.3379... 0.3069 sec/batch
step: 9410/10000... loss: 3.1551... 0.3352 sec/batch
step: 9420/10000... loss: 3.1685... 0.3188 sec/batch
step: 9430/10000... loss: 3.1777... 0.3131 sec/batch
step: 9440/10000... loss: 3.1411... 0.3131 sec/batch
step: 9450/10000... loss: 3.0877... 0.3135 sec/batch
step: 9460/10000... loss: 3.3133... 0.3007 sec/batch
step: 9470/10000... loss: 3.2650... 0.3144 sec/batch
step: 9480/10000... loss: 3.0572... 0.3231 sec/batch
step: 9490/10000... loss: 3.1727... 0.3153 sec/batch
step: 9500/10000... loss: 3.1691... 0.3387 sec/batch
step: 9510/10000... loss: 3.2173... 0.3115 sec/batch
step: 9520/10000... loss: 3.1775... 0.3181 sec/batch
step: 9530/10000... loss: 3.1452... 0.3267 sec/batch
step: 9540/10000... loss: 3.0606... 0.3385 sec/batch
step: 9550/10000... loss: 3.2293... 0.3061 sec/batch
step: 9560/10000... loss: 3.2285... 0.3266 sec/batch
step: 9570/10000... loss: 3.1815... 0.3187 sec/batch
step: 9580/10000... loss: 3.1902... 0.3049 sec/batch
step: 9590/10000... loss: 3.1574... 0.3221 sec/batch
step: 9600/10000... loss: 3.2022... 0.3164 sec/batch
step: 9610/10000... loss: 3.2311... 0.3395 sec/batch
step: 9620/10000... loss: 3.2575... 0.3259 sec/batch
step: 9630/10000... loss: 3.0658... 0.3335 sec/batch
step: 9640/10000... loss: 3.1704... 0.3184 sec/batch
step: 9650/10000... loss: 3.1346... 0.3250 sec/batch
step: 9660/10000... loss: 3.2999... 0.3110 sec/batch
step: 9670/10000... loss: 3.1509... 0.3238 sec/batch
step: 9680/10000... loss: 3.2025... 0.3209 sec/batch
step: 9690/10000... loss: 3.1963... 0.3092 sec/batch
step: 9700/10000... loss: 3.1393... 0.3354 sec/batch
step: 9710/10000... loss: 3.0508... 0.3166 sec/batch
step: 9720/10000... loss: 3.2764... 0.3266 sec/batch
step: 9730/10000... loss: 3.3006... 0.3335 sec/batch
step: 9740/10000... loss: 3.0541... 0.3164 sec/batch
step: 9750/10000... loss: 3.1417... 0.3032 sec/batch
step: 9760/10000... loss: 3.1056... 0.3080 sec/batch
step: 9770/10000... loss: 3.2394... 0.3289 sec/batch
step: 9780/10000... loss: 3.1666... 0.3105 sec/batch
step: 9790/10000... loss: 3.2028... 0.3227 sec/batch
step: 9800/10000... loss: 3.0550... 0.3166 sec/batch
step: 9810/10000... loss: 3.2708... 0.3204 sec/batch
step: 9820/10000... loss: 3.2640... 0.3010 sec/batch
step: 9830/10000... loss: 3.2557... 0.3160 sec/batch
step: 9840/10000... loss: 3.1933... 0.3166 sec/batch
step: 9850/10000... loss: 3.1658... 0.3146 sec/batch
step: 9860/10000... loss: 3.1742... 0.3244 sec/batch
step: 9870/10000... loss: 3.2502... 0.3032 sec/batch
step: 9880/10000... loss: 3.2122... 0.3240 sec/batch
step: 9890/10000... loss: 3.0794... 0.3075 sec/batch
step: 9900/10000... loss: 3.1849... 0.3255 sec/batch
step: 9910/10000... loss: 3.1823... 0.3157 sec/batch
step: 9920/10000... loss: 3.3323... 0.3082 sec/batch
step: 9930/10000... loss: 3.1534... 0.3184 sec/batch
step: 9940/10000... loss: 3.1557... 0.3174 sec/batch
step: 9950/10000... loss: 3.1723... 0.3199 sec/batch
step: 9960/10000... loss: 3.1096... 0.3080 sec/batch
step: 9970/10000... loss: 3.0905... 0.3286 sec/batch
step: 9980/10000... loss: 3.2738... 0.3113 sec/batch
step: 9990/10000... loss: 3.2860... 0.3190 sec/batch
step: 10000/10000... loss: 3.0485... 0.3190 sec/batch
In [14]:
!cd Char-RNN-TensorFlow && python sample.py --converter_path model/writings/converter.pkl --checkpoint_path model/writings --use_embedding --max_length 2000 --num_layers 3 --lstm_size 256 --embedding_size 256
WARNING:tensorflow:From /content/Char-RNN-TensorFlow/model.py:93: softmax_cross_entropy_with_logits (from tensorflow.python.ops.nn_ops) is deprecated and will be removed in a future version.
Instructions for updating:
Future major versions of TensorFlow will allow gradients to flow
into the labels input on backprop by default.
See @{tf.nn.softmax_cross_entropy_with_logits_v2}.
2018-07-28 15:51:34.229249: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:897] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2018-07-28 15:51:34.229797: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1392] Found device 0 with properties:
name: Tesla K80 major: 3 minor: 7 memoryClockRate(GHz): 0.8235
pciBusID: 0000:00:04.0
totalMemory: 11.17GiB freeMemory: 11.10GiB
2018-07-28 15:51:34.229845: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1471] Adding visible gpu devices: 0
2018-07-28 15:51:34.641848: I tensorflow/core/common_runtime/gpu/gpu_device.cc:952] Device interconnect StreamExecutor with strength 1 edge matrix:
2018-07-28 15:51:34.641926: I tensorflow/core/common_runtime/gpu/gpu_device.cc:958] 0
2018-07-28 15:51:34.641953: I tensorflow/core/common_runtime/gpu/gpu_device.cc:971] 0: N
2018-07-28 15:51:34.642401: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1084] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 10763 MB memory) -> physical GPU (device: 0, name: Tesla K80, pci bus id: 0000:00:04.0, compute capability: 3.7)
Restored from: model/writings/model-10000
<unk>，我们不愿意面对我自己不愿意面对自己有时光。他们是一种一切的一切，我们不能在别人懂得自己的生命状态和自己，而我们的意愿是在我们自己的意愿，在这个世界无限之后的时间，他们的人生就是在心的意识成长来的时候，就不能使你感觉到我的心理时。
我们在自己想得的东西都不过是一种人类的时候，就会使这种复杂的空，这些事情不过是一个人，而不会被蜿蜒的心情和平衡。
在一种语言中发现了一点理想主张的人，就是对于人类人类历史时候。
人生的是这种意识形态的发展是我生活的特意，不过是一个不能忍受一个人，可以用来做我人的一个人。人们不再在一个个人都有些人们的生命，我们不要在自己的生产状态，或许自己为了一些不愿意的人生都会不会以一个个一个主观时一步的人的事情，可这一个人的心，我的一个基本消灭，在自己孜孜可以不可能做得一种无垠。我们不再有一个人的心，可是在我们不断成为一个人的主体和客宰在的人们。我们不是我们的信仰和我们自己的自己，那些我们对自己的生命，是有人的意义是否是自己的生命和人们生活的夹料，在我们的心情一直就是我的一种优先方，我们在一个世界无法做成，也没有一种无疑，而是我们自己的意志和心实生活，我们不能在自己的人生和自己。我们对我的心理平衡点不是一个人的主张的基础，是我们的意志与心理生存，这一切就是自己的心理平衡，可是在你们自身地做到这个世界的时候，又没有一个人的人，可在于这个世界的人生，我们在一种自我辩护的人，在自己的生存状态和人生，那么我只是这种有限的人，我要不愿意面对这个事情的时间，不过会不是一样的心情。我们在这世界上要做一个人的人，可不过是一种人的需要我们在一个人都是我们的心中。
在此一个人的一个时间里，我们的人生的人类在心中无法不断，可是我们不会做到的事件，不是因为这些自己的生命和无力；我们的思维不可能鼓缩了这种道理、真相、自己的意识形态不同的态度。我们不要忘记一个自己，害怕一个自己的意义，我们不要揭坏自己的心灵和平衡、丑恶和我自己和自己的生命。
人类对我生存性生命的平衡和一个自觉，在我们不知意做了自己的自己，而不是一个人，不能在人们留下了不断的事情。
如果我们想要自己的一个信仰，我们对这个世界的人生，是一个有小的事情。然而我们的信仰和自己的生存，就有一个人在自身渗灭了，是我们不知道这一点。
不能
----
我们看见一种思维，仔细的时候是一个自己辩护的人们的思想和一个对象的人的事情和实在，而不是因为它们对这个世界的基础，而是一个有法律的人。人们在这种世界上不过是用了人的存在方式，是那种一种一个个人的一个人。人们在一个碳不可靠的牛因人的理解。我们在这个世界里成为一个人的人的事情，我们可以接受一个人的世界。一个人的人的心，一个人的心，可在我们在一个人似样地做的，是我们一个人在自己的生命力，我们的主体的主体和自己在自己的自己和自己的世界无力和心情的真实，我们可以在人类历史一些自己力量的方式，不能被实现方一个主体）的存在浓缩着这个世界的一切；我们就能在人类在人生中不可能被迫到人生命中时的一个人在人格之中一种人类的意志和实际性的存在的意识形态和一个人类的基础和历史，使你们在人格的自我，然而这样，他的心理中无法做了，又可以不可挡用我们自己的方式不能成为我们的意愿。我们在一种无数湮亡的一面，我的心灵不能不可抗拒的人生都会放弃对我生活的意志。
一个人不能调成自己的生存状态，我的意义不是一个一种一个个人的事情；一种人的世界，不能在自己的世界里，我也会不会做成自己的世界，我们不再是“爱所有的事情。
如此一个人在一个人在一切地做的是人的人。但那个人们在一个人的双方和应该有机制的时间，我们就会在这一个个人只有这种世界的时候，我会在这世界不可能把人类留起了无限和研始，可是一个有人之间不过是自己的世界。一切不能不能解决自己。
我是一个人在人生的一面。在自己的意识形态和基石，不过是用人生的理想，篡这个人的心情一定会在一种自己，因为我们的意愿不能及自己在自己和自己的心理平衡点，可是深陷了不可能地变得臃燥乏力。
如果我们不能说那种世界不能保定这个世界。
我们要看见自己的时候，我们都像这个字，一个人的关键，我们在我们的意愿不是在一个人类生活中成为一个人类的主体的基础和资本和一点，我在这个世界里成为一个主体的期望，这些人们在这种人生的生产力量的一种基本性。可是我们被迫在这些世界上的人们，我们在我们的心理挣扎，不可是是自己不愿意的人们在于自己的意义和意义，而我们在我们的心情平静也不是一样，而且我们不会看着自己的时间去做了，这是一种无法自然的自己，我们还是一个人的人的事情和一个人，有的碎片会在你们做了一个有自由的一个副产生。
我们对我们被文本和人类的生活状态，在人们的人生，是我们的意义，我们还是一个人的意义。
在这个时候，我就是一种不喜欢的事情，在我们的心中
In [15]:
!cd Char-RNN-TensorFlow && python sample.py --converter_path model/writings/converter.pkl --checkpoint_path model/writings --use_embedding --max_length 2000 --num_layers 3 --lstm_size 256 --embedding_size 256
WARNING:tensorflow:From /content/Char-RNN-TensorFlow/model.py:93: softmax_cross_entropy_with_logits (from tensorflow.python.ops.nn_ops) is deprecated and will be removed in a future version.
Instructions for updating:
Future major versions of TensorFlow will allow gradients to flow
into the labels input on backprop by default.
See @{tf.nn.softmax_cross_entropy_with_logits_v2}.
2018-07-28 15:55:27.711992: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:897] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2018-07-28 15:55:27.712659: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1392] Found device 0 with properties:
name: Tesla K80 major: 3 minor: 7 memoryClockRate(GHz): 0.8235
pciBusID: 0000:00:04.0
totalMemory: 11.17GiB freeMemory: 11.10GiB
2018-07-28 15:55:27.712720: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1471] Adding visible gpu devices: 0
2018-07-28 15:55:28.291497: I tensorflow/core/common_runtime/gpu/gpu_device.cc:952] Device interconnect StreamExecutor with strength 1 edge matrix:
2018-07-28 15:55:28.291580: I tensorflow/core/common_runtime/gpu/gpu_device.cc:958] 0
2018-07-28 15:55:28.291611: I tensorflow/core/common_runtime/gpu/gpu_device.cc:971] 0: N
2018-07-28 15:55:28.292032: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1084] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 10763 MB memory) -> physical GPU (device: 0, name: Tesla K80, pci bus id: 0000:00:04.0, compute capability: 3.7)
Restored from: model/writings/model-10000
况远的东西。我是一个人在自己一个人的心情中的事情一直会成为一种人类历史时间的发展，不是一个人的一个小人，不是不过是理想的事情。』
摘抄来源：http://www.zouban.com/neting/nevanitic/209962912949/
http://book.doubing.com/rever/2836999321/
阴谋论
-----
> 『在我们不可怕的，所有的人都要不会没有人生，而没有做到自己，但我们不会被因为自己，害怕我们，在自己的生存状态。我们对自由一点执着的时候，不会不过是一种自由和自由和孤寂地使自己的意愿，在一个人的心理时候就会不能重复一件人，一切都是自由，不会是一种不可不能合心的一个副命境的，只有当时的平衡点去有了不可能的。』
摘抄来源：http://book.douban.com/note/123446669967533/
http://www.douban.com/notiew/274619/]hom/ronaliting/13939/
http://baca.com/revilw//www.doubin.com/note/1639662346039339/
排序
----
> 『人呵，“我自己不再是自己的生活力量的一个毒瘤。而是我们各一个事，很好，是人们不能懂得自己的一个主观，不是在人类来做了自己和人类的一切对于自我不能自由。而我们的人生的自己在这个主体的一个基本体在，是一种政治时候，在我们的人生都无法抗拒这个社会，不过是自己的世界。
在某种人类在人的信图保护自己，这些一点不能在人生中涓涓的滋环。
我们在我们的生产力的桎梏，在人生的文字中不断到一个一种不祥的火焰。
我们不能看见一个痴情，在人生的人生的意义里不能成为我们所有的世界，那些我们的心中一定会在自己的心灵平衡的心情和习准和不断上的，只能一种带不到人的人。
如果神有一种不平地。
我们在一起不是我，我们在自己想想的事件中，我们不愿意做这种人们，可是一定会有一种可能的人的事情。我们在这些世界上成为这些意愿形态的时候，我们不会让它看到一个东西，可是在这一种人性里的一点。这种世界，在我们对这种不可遏制的人，不能因为别人的意义和在于人类生活中无法消灭和人的人，人们的一个基本消灭，我们还是会做到这些一种人类的存在。
我们对于每个主体时间，我们还是在自己的人生和巨大而徒破人地的人，这种不可承担的人，在自己不能看出自己的世界，在人的意义在一切可能成为的一个事情，我们在这种世界不能被放弃。
人们的文源是不断上升我们心理学，不能不是因为他的心灵时间。
我们将一起被灌输来地做自己的人，在自己的人生，不过是自己的生命，我们是这个世界。我们是一个人的人，我们的长辈们，只有一个平衡的人们，只是我们有一种人类的生存，我们在我们的意义和价值观意成为自己不能做出的。他们在一个世界里不断改变的事情，都会不能被蜿蜒的，是人类的心理平衡点；在这个世界示快的一种不平的人？这一点的是人的人，是人类为人生中的人生的一个人的事情的时间，我们不能做了。
我的誓气锋常，是一个不能做的事情。我们徜徉在于一个主体的基分。人生的人们在于一个人的人的世界。这个人们的梦想，我不能使这种世界上的一切砖上。
这样的话，我们在一种自我心念生活的轰轰丛烈》和人生，一切会看到这个事情的事情。我们被放弃，我们不会不能做来，我们不能让人们做起了自觉的自己。
人们的思维来不清楚。我们一个人在人类和人生，不过是一个人的心。一切都是为什么不可以不过是一种人生的生活。
我在人生的人生和一切，在一个人修着自己的人们，无法摆脱自己的生存，我们不能抵报我们心情的心情。
我不愿意只有一切可以让我们的思想自由一种被不可能性。这是一种被美术的栖居、不得和自己的力量的时候，我们的心脏衰弱起了我们不愿白了我们，不愿意始终不能因为一个慎重的世界和自由和自由的世界，我们是一个不是一种不祥的态量，使人不可以被改变一个人生的时间变得不断的路途。而且，这是一种绝望的不能使一个人在世界的一个人的世界，我们在这个世界里不是这种贴则，从而心灵不足地无限。但是我们在自己所能做的，是一种无力的事情，也是一个有限的时候，我已经能够做起来。
这样的一切就是一个人，是人们无法做成。所以，我们不是在自觉的人格。我们的焦虑与恐惧不能懂得。
如果没有勇气，就是一种不可承担，因为它们不过是一起自己的心中不可以做到的地方。
我们在自己的心中平等和自己的意识在于这种不可能的人们，然后这种不是我的人，是否没有自己的愿望和不可能。因为你不能不能改变我们对于人类人格的人格。我们不能看见这种不能忍受了我们的心理时光，就不能是这些烂泥塘的碎片一样的心情。
我们将不是自己和不愿意的自己和自己，不会不是这样的一个小小，一种无法不断地不能重新
In [17]:
!cd Char-RNN-TensorFlow && python sample.py --converter_path model/writings/converter.pkl --checkpoint_path model/writings --use_embedding --max_length 2000 --num_layers 3 --lstm_size 256 --embedding_size 256
WARNING:tensorflow:From /content/Char-RNN-TensorFlow/model.py:93: softmax_cross_entropy_with_logits (from tensorflow.python.ops.nn_ops) is deprecated and will be removed in a future version.
Instructions for updating:
Future major versions of TensorFlow will allow gradients to flow
into the labels input on backprop by default.
See @{tf.nn.softmax_cross_entropy_with_logits_v2}.
2018-07-28 15:56:24.743002: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:897] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2018-07-28 15:56:24.743732: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1392] Found device 0 with properties:
name: Tesla K80 major: 3 minor: 7 memoryClockRate(GHz): 0.8235
pciBusID: 0000:00:04.0
totalMemory: 11.17GiB freeMemory: 11.10GiB
2018-07-28 15:56:24.743790: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1471] Adding visible gpu devices: 0
2018-07-28 15:56:25.285580: I tensorflow/core/common_runtime/gpu/gpu_device.cc:952] Device interconnect StreamExecutor with strength 1 edge matrix:
2018-07-28 15:56:25.285684: I tensorflow/core/common_runtime/gpu/gpu_device.cc:958] 0
2018-07-28 15:56:25.285716: I tensorflow/core/common_runtime/gpu/gpu_device.cc:971] 0: N
2018-07-28 15:56:25.286206: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1084] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 10763 MB memory) -> physical GPU (device: 0, name: Tesla K80, pci bus id: 0000:00:04.0, compute capability: 3.7)
Restored from: model/writings/model-10000
睛
- 惟有什么我不想做，只是因为这一种不能是一种不可能的事情，都是在这个世界的事情，不会有时候，有的人们在一个人自己的心理平衡点去不得不能重复自己，这些世界的真正历史和不够自己的自由，而这是我们，只有自己，这是一样的事情，都是在这种过程中，不能因为这种不知是一些自觉的一个基本中在一个人在个人的人，是自己不在卑水的时间，都有一种无法，在人们的意义就被在这些世界上做了。』
我在世界的一个角度，这是一个不可逾越的人，在我们对自己的自己。
我不愿再做这个社会的存在。我们不再是一种“大””。
我们对这个人，我们怎么不想在这个世界，不过是不是一切“破坏的心，不能被破坏。
一种亢趣，不是理论，有一个词是一定有人生的人，不能不可以改变演成一种不断的火潭。这是一个个人都不能不过，不是我们是人的一种淡漠的时间。
人与世界，在这些世界里成为一个人的人。我们在自己所做的，所以我们的心理无法，我们不会再做人们做出了的人，我们还是一些人的人的人。我们在一切不能汲取力量和这个主体，我们还是在人类共同界，以为这个世界里，这样的一个孩子，是那种不可挽用的，另外一个人头都攫取和动力的机制，而无疑是有据一定之间，才会有一个人的一切的心理人，这个一个主体的一件人的智慧，是一种自我辩护，不能因为自己的意愿。
我们在我们的思维依附一个正常的人，一个人对于这样的人生都是无力的事情，不是一个人，而是自己的人生就是无数者不能不断，可以不是我们对自己的自我不愿，我们的心灵无力的地方稍稍一点就被一个人的心，一个是人生的一面。我们不再不能看到我们的信点和一切，我们不要不能在我们在自己所有的初衷，我们在自己，做到了我们的心灵的心。我们在我们的思考，我们不会再做自己的生存。
这些人，我的心情无法不清楚，我们的信念和一种淡淡。这种不是诗。我们不能抵怨一切淡泊的心情，让人们在人生里的一面。人的生命无法不可断在这个世界的人生，我们不能做的，是我们自己的意志，是一个自我不断地做到。我是一个不能踩心中的时候，它让人做得不断的态度，在一个人的一个人的事情发展了我们的心灵时间的心理时间。』
- 人类的人们，在一个主体主义的意义和二分、人生的一个基本体系。一种能力是一种完美的基分的时代，就有这么一个一种人，是人类人生的意思是在人类人生的意识形态不同，可能以不够及人类在一个人”的一个主体），人们社会的时间就会不会被蜿蜒的时间里，在自己孜孜不是一些一种一种莫名的人生，而是关于自身的人们，不过不过是一个自由的心理阴影。我们是不要做出的事情这种不是一种不祥的态度。
我们将自身渗入人，这个世界，有的人不能看见自己的世界观本为这种不平的，我们的内心则是一个有小小的阅读中和世界，我们在这些烂泥塘中继续不断不断的路途。
在这个社会的基本上带来，如果有些人的错误就有了一个不足的人，在一种不断和人的人。人们被放着这种多么时间里，就是一种不可断。
我们不要说，我们在自己的意志在一个人似一个小人在一个人在一个世界串事情的事情，不过是不是我们的心中，不可能在一些人。今年是我的生活。我们在我不能做出的，我们还是在自己的心心平衡，不是因为这样的一切不可以不能做得自己，而不是在自己在自己的本质和自由，抑至是在一个人的无法做到，让自己生活和不可以做的事情，不能不可以改变一个人的人。
我们在这个世界之间不是一种不断的态度，是一种被触角”，不能因为自己，他们会会发现自己不会不能想得自由这一点，都像这样不是一切。我们有时生活于一切，从而不能被无法袒露的函法。
我们在劫地被无法不可忍担，不能使自己在人生的自由），可以失去了一些自己的意愿。
我不愿再是自己的生活，我们不愿意不能揣行到那种心理摔起一个人的世界，在我们对这些世界里的存在，是那个世界上的，只能不可挡地不过是自己的心理一种不是在冥理，我们在一个主体内心无法不断地回到，我们的心心旅将一直不能不能及自由人生的一种小小的时候，会有自身面对自己，我们不再不能做你的自己。
这些人，是我不能做的。我想要做的一切，我们不会在自己的自己，我们是这个不是我的愿望，我们在这个世界上成为人类丑恶的人，我们不愿意为我们成为我们自由的人？
这种人生的意义就是在一种无法可断的东西，我们的人生，是一种自我辩境的原因，是那个不可能的。这样的时候，我们都会看见自己的心理平衡，是一个人们都会在于这些世界里的时候，我们的人生的自我，我们在这些社会在人格的时候，不是我们的意义和诠决，我们还会能做自己自由。我们的人们无法做出，我们的信仰和我们一个有人生的人，是人生的人，只能不再能做到事情的智慧。这是迄今不可靠的事情，也不是真的无限。
一切的一个人，在我的心灵看起了我们自己的心情，我们才要拆开和自由。我们不再不断向自己，不过在自己的心理平衡点不过是自己的自己，而是在自己
In [0]:
Content source: utensil/julia-playground
Similar notebooks: