min_char_RNN
a simple character RNN implement.
statement
- The lost of source code was wrote according to karpathy's gist
- You can get background knowledge from here
what's this
- These code will training a rnn mode to generate article.
- A mode file will be created when you kill the process(eg: Ctrl+C).
- You can load pre-training mode to do a generate test.
how to use
- run
python training.py
to do a training. An example output is like below:
iter 2000, loss: 70.034620
iter 2100, loss: 69.157650
iter 2200, loss: 68.066171
iter 2300, loss: 67.368618
iter 2400, loss: 66.422001
iter 2500, loss: 65.549434
iter 2600, loss: 64.841401
iter 2700, loss: 64.127791
iter 2800, loss: 63.122929
iter 2900, loss: 62.588574
generator output:
re pnome.
-all whelld beage pit theel pepore che pnebeit be the the nlod woror mat ll latkers fotiy soet
'lo bs
Touper theounn chas hef bure
I but of and teein tuea.:
Whaild faglait bard masr aseind C
iter 3000, loss: 62.092721
- run
python generator.py
will load model file to generate string.