Readme

Description

This includes a very small start at abstract 'dataset' and 'bot' classes, a quick / dirty implementation of a seq2seq bot, an 'echo bot' which just repeats the user's words, and a data pipeline for the Cornell movie dialog dataset.

Everything will change, this is just fun / starting playing around.

The seq2seq bot and data pipeline are mostly ripped from the Tensorflow fr-en translation example located here.

Requires:

Tensorflow

Internet connection for python to download the data

Usage:

Echo Bot:

Interactive testing / chat:

python echo_bot.py

Seq2Seq Bot:

Training:

python seq2seq_bot.py

Interactive testing / chat:

python seq2seq_bot.py --converse=True

Config:

See seq2seq_bot.py for available flags (e.g. layers, units, batch size) and defaults.

Notes:

It will train very slowly on a laptop. Better to use a GPU server for all but debugging.

On Han Liu's Tesla K40c server, it looks like train time will be on the order of days, depending on # layers etc.

Sample conversation trace

After ~100,000 train steps (~8 hours) - should probably train for ~3 days on this dataset.

(tensorflow)[abeatson@smile simple_bot]$ python seq2seq_bot.py --reload=True
I tensorflow/stream_executor/dso_loader.cc:111] successfully opened CUDA library libcublas.so locally
I tensorflow/stream_executor/dso_loader.cc:111] successfully opened CUDA library libcudnn.so locally
I tensorflow/stream_executor/dso_loader.cc:111] successfully opened CUDA library libcufft.so locally
I tensorflow/stream_executor/dso_loader.cc:111] successfully opened CUDA library libcuda.so.1 locally
I tensorflow/stream_executor/dso_loader.cc:111] successfully opened CUDA library libcurand.so locally
I tensorflow/core/common_runtime/gpu/gpu_device.cc:951] Found device 0 with properties:
name: Tesla K40c
major: 3 minor: 5 memoryClockRate (GHz) 0.745
pciBusID 0000:81:00.0
Total memory: 11.17GiB
Free memory: 11.10GiB
I tensorflow/core/common_runtime/gpu/gpu_device.cc:972] DMA: 0
I tensorflow/core/common_runtime/gpu/gpu_device.cc:982] 0:   Y
I tensorflow/core/common_runtime/gpu/gpu_device.cc:1041] Creating TensorFlow device (/gpu:0) -> (device: 0, name: Tesla K40c, pci bus id: 0000:81:00.0)
Reading model parameters from tmp/bot_train/translate.ckpt-95400
> Hello
Hello .
> What is your name ?
Uh huh ?? William name does .
> Who are you?
My name is _UNK .
> What do you do?
I ' m retired . . .
> That's nice
Yeah .
> Tell me a story
Uh . . . uh . . .
> You're boring
I ' m not .
> How is your day today?
Oh , my good .
> Goodbye
Goodbye .