-
Notifications
You must be signed in to change notification settings - Fork 42
/
Copy pathtraining_smiles_language_model_from_scratch.html
183 lines (141 loc) · 9.94 KB
/
training_smiles_language_model_from_scratch.html
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
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
---
title: Train a SMILES language model from scratch
keywords: fastai
sidebar: home_sidebar
summary: "Tutorial how to train a reaction language model"
description: "Tutorial how to train a reaction language model"
nb_path: "nbs/08_training_smiles_language_model_from_scratch.ipynb"
---
<!--
#################################################
### THIS FILE WAS AUTOGENERATED! DO NOT EDIT! ###
#################################################
# file to edit: nbs/08_training_smiles_language_model_from_scratch.ipynb
# command to build the docs after a change: nbdev_build_docs
-->
<div class="container" id="notebook-container">
{% raw %}
<div class="cell border-box-sizing code_cell rendered">
</div>
{% endraw %}
{% raw %}
<div class="cell border-box-sizing code_cell rendered">
<div class="input">
<div class="inner_cell">
<div class="input_area">
<div class=" highlight hl-ipython3"><pre><span></span><span class="kn">import</span> <span class="nn">os</span>
<span class="kn">import</span> <span class="nn">numpy</span> <span class="k">as</span> <span class="nn">np</span>
<span class="kn">import</span> <span class="nn">pandas</span> <span class="k">as</span> <span class="nn">pd</span>
<span class="kn">import</span> <span class="nn">torch</span>
<span class="kn">import</span> <span class="nn">logging</span>
<span class="kn">import</span> <span class="nn">random</span>
<span class="kn">from</span> <span class="nn">rxnfp.models</span> <span class="kn">import</span> <span class="n">SmilesLanguageModelingModel</span>
<span class="n">logger</span> <span class="o">=</span> <span class="n">logging</span><span class="o">.</span><span class="n">getLogger</span><span class="p">(</span><span class="vm">__name__</span><span class="p">)</span>
</pre></div>
</div>
</div>
</div>
</div>
{% endraw %}
<div class="cell border-box-sizing text_cell rendered"><div class="inner_cell">
<div class="text_cell_render border-box-sizing rendered_html">
<h2 id="Track-the-training">Track the training<a class="anchor-link" href="#Track-the-training"> </a></h2><p>We will be using wandb to keep track of our training. You can use the an account on <a href="https://www.wandb.com">wandb</a> or create an own instance following the instruction in the <a href="https://docs.wandb.com/self-hosted">documentation</a>.</p>
<p>If you then create an <code>.env</code> file in the root folder and specify the <code>WANDB_API_KEY=</code> (and the <code>WANDB_BASE_URL=</code>), you can use dotenv to load those enviroment variables.</p>
</div>
</div>
</div>
{% raw %}
<div class="cell border-box-sizing code_cell rendered">
<div class="input">
<div class="inner_cell">
<div class="input_area">
<div class=" highlight hl-ipython3"><pre><span></span><span class="c1"># !pip install python-dotenv</span>
<span class="kn">from</span> <span class="nn">dotenv</span> <span class="kn">import</span> <span class="n">load_dotenv</span><span class="p">,</span> <span class="n">find_dotenv</span>
<span class="n">load_dotenv</span><span class="p">(</span><span class="n">find_dotenv</span><span class="p">())</span>
</pre></div>
</div>
</div>
</div>
</div>
{% endraw %}
<div class="cell border-box-sizing text_cell rendered"><div class="inner_cell">
<div class="text_cell_render border-box-sizing rendered_html">
<h2 id="Setup-MLM-training">Setup MLM training<a class="anchor-link" href="#Setup-MLM-training"> </a></h2><p>Choose the hyperparameters you want and start the training. The default parameters will train a BERT model with 12 layers and 4 attention heads per layer. The training task is Masked Language Modeling (MLM), where tokens from the input reactions are randomly masked and predicted by the model given the context.</p>
<p>After defining the config, the training is launched in 3 lines of code using our adapter written for the <a href="https://simpletransformers.ai">SimpleTransformers</a> library (based on huggingface <a href="https://github.com/huggingface/transformers">Transformers</a>).</p>
<p>To make it work you will have to install simpletransformers:</p>
<div class="highlight"><pre><span></span>pip install simpletransformers
</pre></div>
</div>
</div>
</div>
{% raw %}
<div class="cell border-box-sizing code_cell rendered">
<div class="input">
<div class="inner_cell">
<div class="input_area">
<div class=" highlight hl-ipython3"><pre><span></span><span class="n">config</span> <span class="o">=</span> <span class="p">{</span>
<span class="s2">"architectures"</span><span class="p">:</span> <span class="p">[</span>
<span class="s2">"BertForMaskedLM"</span>
<span class="p">],</span>
<span class="s2">"attention_probs_dropout_prob"</span><span class="p">:</span> <span class="mf">0.1</span><span class="p">,</span>
<span class="s2">"hidden_act"</span><span class="p">:</span> <span class="s2">"gelu"</span><span class="p">,</span>
<span class="s2">"hidden_dropout_prob"</span><span class="p">:</span> <span class="mf">0.1</span><span class="p">,</span>
<span class="s2">"hidden_size"</span><span class="p">:</span> <span class="mi">256</span><span class="p">,</span>
<span class="s2">"initializer_range"</span><span class="p">:</span> <span class="mf">0.02</span><span class="p">,</span>
<span class="s2">"intermediate_size"</span><span class="p">:</span> <span class="mi">512</span><span class="p">,</span>
<span class="s2">"layer_norm_eps"</span><span class="p">:</span> <span class="mf">1e-12</span><span class="p">,</span>
<span class="s2">"max_position_embeddings"</span><span class="p">:</span> <span class="mi">512</span><span class="p">,</span>
<span class="s2">"model_type"</span><span class="p">:</span> <span class="s2">"bert"</span><span class="p">,</span>
<span class="s2">"num_attention_heads"</span><span class="p">:</span> <span class="mi">4</span><span class="p">,</span>
<span class="s2">"num_hidden_layers"</span><span class="p">:</span> <span class="mi">12</span><span class="p">,</span>
<span class="s2">"pad_token_id"</span><span class="p">:</span> <span class="mi">0</span><span class="p">,</span>
<span class="s2">"type_vocab_size"</span><span class="p">:</span> <span class="mi">2</span><span class="p">,</span>
<span class="p">}</span>
<span class="n">vocab_path</span> <span class="o">=</span> <span class="s1">'../data/uspto_1k_TPL/individual_files/vocab.txt'</span>
<span class="n">args</span> <span class="o">=</span> <span class="p">{</span><span class="s1">'config'</span><span class="p">:</span> <span class="n">config</span><span class="p">,</span>
<span class="s1">'vocab_path'</span><span class="p">:</span> <span class="n">vocab_path</span><span class="p">,</span>
<span class="s1">'wandb_project'</span><span class="p">:</span> <span class="s1">'uspto_mlm_temp_1000'</span><span class="p">,</span>
<span class="s1">'train_batch_size'</span><span class="p">:</span> <span class="mi">32</span><span class="p">,</span>
<span class="s1">'manual_seed'</span><span class="p">:</span> <span class="mi">42</span><span class="p">,</span>
<span class="s2">"fp16"</span><span class="p">:</span> <span class="kc">False</span><span class="p">,</span>
<span class="s2">"num_train_epochs"</span><span class="p">:</span> <span class="mi">50</span><span class="p">,</span>
<span class="s1">'max_seq_length'</span><span class="p">:</span> <span class="mi">256</span><span class="p">,</span>
<span class="s1">'evaluate_during_training'</span><span class="p">:</span> <span class="kc">True</span><span class="p">,</span>
<span class="s1">'overwrite_output_dir'</span><span class="p">:</span> <span class="kc">True</span><span class="p">,</span>
<span class="s1">'output_dir'</span><span class="p">:</span> <span class="s1">'../out/bert_mlm_1k_tpl'</span><span class="p">,</span>
<span class="s1">'learning_rate'</span><span class="p">:</span> <span class="mf">1e-4</span>
<span class="p">}</span>
</pre></div>
</div>
</div>
</div>
</div>
{% endraw %}
{% raw %}
<div class="cell border-box-sizing code_cell rendered">
<div class="input">
<div class="inner_cell">
<div class="input_area">
<div class=" highlight hl-ipython3"><pre><span></span><span class="n">model</span> <span class="o">=</span> <span class="n">SmilesLanguageModelingModel</span><span class="p">(</span><span class="n">model_type</span><span class="o">=</span><span class="s1">'bert'</span><span class="p">,</span> <span class="n">model_name</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span> <span class="n">args</span><span class="o">=</span><span class="n">args</span><span class="p">)</span>
</pre></div>
</div>
</div>
</div>
</div>
{% endraw %}
{% raw %}
<div class="cell border-box-sizing code_cell rendered">
<div class="input">
<div class="inner_cell">
<div class="input_area">
<div class=" highlight hl-ipython3"><pre><span></span><span class="c1"># !unzip ../data/uspto_1k_TPL/individual_files/mlm_training.zip -d ../data/uspto_1k_TPL/individual_files/</span>
<span class="n">train_file</span> <span class="o">=</span> <span class="s1">'../data/uspto_1k_TPL/individual_files/mlm_train_file.txt'</span>
<span class="n">eval_file</span> <span class="o">=</span> <span class="s1">'../data/uspto_1k_TPL/individual_files/mlm_eval_file_1k.txt'</span>
<span class="n">model</span><span class="o">.</span><span class="n">train_model</span><span class="p">(</span><span class="n">train_file</span><span class="o">=</span><span class="n">train_file</span><span class="p">,</span> <span class="n">eval_file</span><span class="o">=</span><span class="n">eval_file</span><span class="p">)</span>
</pre></div>
</div>
</div>
</div>
</div>
{% endraw %}
</div>