-
Notifications
You must be signed in to change notification settings - Fork 113
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
- Loading branch information
0 parents
commit 06b18b4
Showing
10 changed files
with
1,006 additions
and
0 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,44 @@ | ||
data/* | ||
log/* | ||
*.log | ||
|
||
# Compiled Lua sources | ||
luac.out | ||
|
||
# luarocks build files | ||
*.src.rock | ||
*.zip | ||
*.tar.gz | ||
|
||
# Object files | ||
*.o | ||
*.os | ||
*.ko | ||
*.obj | ||
*.elf | ||
|
||
# Precompiled Headers | ||
*.gch | ||
*.pch | ||
|
||
# Libraries | ||
*.lib | ||
*.a | ||
*.la | ||
*.lo | ||
*.def | ||
*.exp | ||
|
||
# Shared objects (inc. Windows DLLs) | ||
*.dll | ||
*.so | ||
*.so.* | ||
*.dylib | ||
|
||
# Executables | ||
*.exe | ||
*.out | ||
*.app | ||
*.i*86 | ||
*.x86_64 | ||
*.hex |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,140 @@ | ||
# Neural Conversational Model in Torch | ||
|
||
This is an attempt at implementing [Sequence to Sequence Learning with Neural Networks (seq2seq)](http://arxiv.org/abs/1409.3215) and reproducing the results in [A Neural Conversational Model](http://arxiv.org/abs/1506.05869) (aka the Google chatbot). | ||
|
||
The Google chatbot paper [became famous](http://www.sciencealert.com/google-s-ai-bot-thinks-the-purpose-of-life-is-to-live-forever) after cleverly answering a few philosophical questions, such as: | ||
|
||
> **Human:** What is the purpose of living? | ||
> **Machine:** To live forever. | ||
## How it works | ||
|
||
The model is based on two [LSTM](https://en.wikipedia.org/wiki/Long_short-term_memory) layers. One for encoding the input sentence into a "thought vector", and another for decoding that vector into a response. This model is called Sequence-to-sequence or seq2seq. | ||
|
||
![seq2seq](https://4.bp.blogspot.com/-aArS0l1pjHQ/Vjj71pKAaEI/AAAAAAAAAxE/Nvy1FSbD_Vs/s640/2TFstaticgraphic_alt-01.png) | ||
_Source: http://googleresearch.blogspot.ca/2015/11/computer-respond-to-this-email.html_ | ||
|
||
In this experiment, we train the seq2seq model with movie dialogs from the [Cornell Movie-Dialogs Corpus](http://www.mpi-sws.org/~cristian/Cornell_Movie-Dialogs_Corpus.html). The lines are shortened to the first sentence. | ||
|
||
## Sample conversation | ||
|
||
Here's a sample conversation after training for 20 epoch with 50000 examples, using the following command: | ||
|
||
```sh | ||
th train.lua --cuda --dataset 50000 --hiddenSize 1000 | ||
``` | ||
|
||
(Took 3 days to train on my GeForce GTX 780M.) | ||
|
||
For OpenCL, use `--opencl` instead of `--cuda`. To train on CPU, don't provide any of those two. | ||
|
||
> **me:** Hello? | ||
> **bot:** Hi. | ||
> | ||
> **me:** How are you? | ||
> **bot:** I'm fine. | ||
> | ||
> **me:** What's your name? | ||
> **bot:** It's hard to describe. | ||
> | ||
> **me:** How so? | ||
> **bot:** I'm not sure. | ||
> | ||
> **me:** What color is the sky? | ||
> **bot:** It's blue. | ||
> | ||
> **me:** What is your job? | ||
> **bot:** It's not that i'm a fucking werewolf! | ||
> | ||
> **me:** What is the purpose of life? | ||
> **bot:** A gift. | ||
> | ||
> **me:** Are you intelligent? | ||
> **bot:** Yes, well... | ||
> | ||
> **me:** Are you a machine? | ||
> **bot:** That's a lie. | ||
> | ||
> **me:** Are you human? | ||
> **bot:** No, i'm not. | ||
> | ||
> **me:** What are you? | ||
> **bot:** I'm not sure. | ||
> | ||
> **me:** Do you plan on taking over the world? | ||
> **bot:** No, i don't. | ||
Phew! That was close. Good thing I didn't train it on the full dataset. Please experiment responsibly. | ||
|
||
_(Disclaimer: nonsensical responses have been removed.)_ | ||
|
||
## Installing | ||
|
||
1. [Install Torch](http://torch.ch/docs/getting-started.html). | ||
2. Install the following additional Lua libs: | ||
|
||
```sh | ||
luarocks install nn | ||
luarocks install rnn | ||
luarocks install penlight | ||
``` | ||
|
||
To train with CUDA install the latest CUDA drivers, toolkit and run: | ||
|
||
```sh | ||
luarocks install cutorch | ||
luarocks install cunn | ||
``` | ||
|
||
To train with opencl install the lastest Opencl torch lib: | ||
|
||
```sh | ||
luarocks install cltorch | ||
luarocks install clnn | ||
``` | ||
|
||
3. Download the [Cornell Movie-Dialogs Corpus](http://www.mpi-sws.org/~cristian/Cornell_Movie-Dialogs_Corpus.html) and extract all the files into data/cornell_movie_dialogs. | ||
|
||
## Training | ||
|
||
```sh | ||
th train.lua [-h / options] | ||
``` | ||
|
||
Use the `--dataset NUMBER` option to control the size of the dataset. Training on the full dataset takes about 5h for a single epoch. | ||
|
||
The model will be saved to `data/model.t7` after each epoch if it has improved (error decreased). | ||
|
||
## Testing | ||
|
||
To load the model and have a conversation: | ||
|
||
```sh | ||
th -i eval.lua --cuda # Skip --cuda if you didn't train with it | ||
# ... | ||
th> say "Hello." | ||
``` | ||
|
||
## License | ||
|
||
MIT License | ||
|
||
Copyright (c) 2016 Marc-Andre Cournoyer | ||
|
||
Permission is hereby granted, free of charge, to any person obtaining a copy | ||
of this software and associated documentation files (the "Software"), to deal | ||
in the Software without restriction, including without limitation the rights | ||
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell | ||
copies of the Software, and to permit persons to whom the Software is | ||
furnished to do so, subject to the following conditions: | ||
|
||
The above copyright notice and this permission notice shall be included in all | ||
copies or substantial portions of the Software. | ||
|
||
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR | ||
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, | ||
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE | ||
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER | ||
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, | ||
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE | ||
SOFTWARE. |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,72 @@ | ||
local CornellMovieDialogs = torch.class("neuralconvo.CornellMovieDialogs") | ||
local stringx = require "pl.stringx" | ||
local xlua = require "xlua" | ||
|
||
local function parsedLines(file, fields) | ||
local f = assert(io.open(file, 'r')) | ||
|
||
return function() | ||
local line = f:read("*line") | ||
|
||
if line == nil then | ||
f:close() | ||
return | ||
end | ||
|
||
local values = stringx.split(line, " +++$+++ ") | ||
local t = {} | ||
|
||
for i,field in ipairs(fields) do | ||
t[field] = values[i] | ||
end | ||
|
||
return t | ||
end | ||
end | ||
|
||
function CornellMovieDialogs:__init(dir) | ||
self.dir = dir | ||
end | ||
|
||
local MOVIE_LINES_FIELDS = {"lineID","characterID","movieID","character","text"} | ||
local MOVIE_CONVERSATIONS_FIELDS = {"character1ID","character2ID","movieID","utteranceIDs"} | ||
local TOTAL_LINES = 387810 | ||
|
||
local function progress(c) | ||
if c % 10000 == 0 then | ||
xlua.progress(c, TOTAL_LINES) | ||
end | ||
end | ||
|
||
function CornellMovieDialogs:load() | ||
local lines = {} | ||
local conversations = {} | ||
local count = 0 | ||
|
||
print("-- Parsing Cornell movie dialogs data set ...") | ||
|
||
for line in parsedLines(self.dir .. "/movie_lines.txt", MOVIE_LINES_FIELDS) do | ||
lines[line.lineID] = line | ||
line.lineID = nil | ||
-- Remove unused fields | ||
line.characterID = nil | ||
line.movieID = nil | ||
count = count + 1 | ||
progress(count) | ||
end | ||
|
||
for conv in parsedLines(self.dir .. "/movie_conversations.txt", MOVIE_CONVERSATIONS_FIELDS) do | ||
local conversation = {} | ||
local lineIDs = stringx.split(conv.utteranceIDs:sub(3, -3), "', '") | ||
for i,lineID in ipairs(lineIDs) do | ||
table.insert(conversation, lines[lineID]) | ||
end | ||
table.insert(conversations, conversation) | ||
count = count + 1 | ||
progress(count) | ||
end | ||
|
||
xlua.progress(TOTAL_LINES, TOTAL_LINES) | ||
|
||
return conversations | ||
end |
Oops, something went wrong.