Decomposable-Attention-Model/INSTRUCTIONS at main · RazNayr/Decomposable-Attention-Model · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Title: Applying Recurrent Layers to the Decomposable Attention Model for Natural Language Inference
Author: Ryan Camilleri
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Instructions
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

1. Setup environment
==============
- Setup python 3.6 environment (conda or any other. Python 3.6 required for tensorflow-gpu)
- Install the following packages:
--> tensorflow-gpu
--> tensorflow
--> keras
--> matplotlib
--> scikit-learn
--> pandas
--> numpy
--> nltk
--> seaborn
--> pydotplus
--> graphviz
--> gensim (If 'ImportError: cannot import name 'open' and using conda', run: 'conda install smart_open==2.0.0')


2. Retrieve SNLI Dataset
==============
- Run data-utils.ipynb


3. Observe Data Distribution
==============
- Open data-visualizer.ipynb
- Run to replicate results


4. Evaluate Models
==============
- Open model-evaluator.ipynb
- Run to replicate results and evaluate the pre-trained models.
(WARNING: This will install the gensim word2vec embeddings if this wasn't already installed)


5. Train Models
==============
- Open model-train.ipynb
- Run to train the defined models.
(WARNING: This will override the pre-trained models that come with this project)
(WARNING: This will install the gensim word2vec embeddings if this wasn't already installed)