forked from minerva-ml/open-solution-toxic-comments
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathneptune.yaml
110 lines (94 loc) · 2.47 KB
/
neptune.yaml
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
project-key: TOX
name: toxic
tags: [glove, lstm, adam]
metric:
channel: 'Final Validation Score ROC_AUC'
goal: maximize
#Comment out if not in Cloud Environment
pip-requirements-file: requirements.txt
exclude:
- output
- imgs
- neptune.log
- offline_job.log
- .git
- .idea
- .ipynb_checkpoints
parameters:
# Cloud Environment
data_dir: /public/toxic_comments
embedding_filepath: /public/models/glove/glove.840B.300d.txt
single_model_predictions_dir: /public/toxic_comments/single_model_predictions_20180226
experiment_dir: /output/trained_pipelines/glove_lstm
# Local Environment
# data_dir: /path/to/toxic/data
# embedding_filepath: /path/to/embedding i.e. ~/glove/glove.840B.300d.txt
# single_model_predictions_dir: /path/to/single/model/preds/ i.e. ~/single_model_predictions_20180226
# experiment_dir: /my/working/directory i.e. ~/toxic/trained_pipelines/glove_lstm
bad_words_filepath: external_data/compiled_bad_words.txt
overwrite: 0
num_workers: 4
n_cv_splits: 5
# Preprocessing
max_features_char: 100
max_features_word: 100000
maxlen_char: 512
maxlen_words: 200
char_ngram_max: 4
drop_punctuation: 1
drop_newline: 1
drop_multispaces: 1
all_lower_case: 1
fill_na_with: ' '
deduplication_threshold: 10
# Architecture
filter_nr: 512
kernel_size: None
repeat_block: 1
dense_size: 256
repeat_dense: 0
max_pooling: 1
mean_pooling: 0
weighted_average_attention: 0
concat_mode: 'concat'
trainable_embedding: 0
word_embedding_size: 300
char_embedding_size: None
# General Architecture
use_prelu: 1
# Log Reg Params
log_reg_c: 4.0
log_reg_penalty: 'l2'
max_iter: 1000
# Ensemble Catboost
catboost__iterations: 500
catboost__learning_rate: 0.02
catboost__depth: 6
catboost__l2_leaf_reg: 10
catboost__border_count: 100
catboost__model_size_reg: None
catboost__rsm: None
catboost__verbose: 0
# Training schedule
epochs_nr: 1000
batch_size_train: 128
batch_size_inference: 128
lr: 0.001
momentum: 0.9
gamma: 0.7
patience: 5
# Regularization
batch_norm_first: 0
use_batch_norm: 0
dropout_embedding: 0.5
rnn_dropout: 0.5
dense_dropout: 0.5
conv_dropout: 0.5
dropout_mode: 'spatial'
rnn_kernel_reg_l2: 0.0001
rnn_recurrent_reg_l2: 0.0001
rnn_bias_reg_l2: 0.0001
dense_kernel_reg_l2: 0.0001
dense_bias_reg_l2: 0.0001
conv_kernel_reg_l2: 0.0001
conv_bias_reg_l2: 0.0001