-
Notifications
You must be signed in to change notification settings - Fork 58
/
Copy pathCHANGES.txt
183 lines (120 loc) · 5.81 KB
/
CHANGES.txt
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
183
v1.2.3 - 2020-05-07
===================
- Updated requirements in order to use newer versions of dependencies
(also fixing potential security vulnerabilities in dependencies)
- Added support for handling model attachments
- Exclamation mark `!` can now be used instead of `__factory__` in
configuration files
v1.2.2.1 - 2019-09-30
=====================
- Added AWS S3 persister
v1.2.2 - 2019-08-15
===================
- Added default Palladium config file locations (to be checked if
`PALLADIUM_CONFIG` is not set)
- Introduced `palladium.dataset.CSV` to replace `palladium.dataset.Table`
- `__copy__` configuration directive now accepts a `__default__`
- Updated requirements in order to use newer versions of dependencies
(also fixing potential security vulnerabilities in dependencies)
v1.2.1.1 - 2018-12-13
=====================
- Added pld-export command to export a model from one model persister
to another.
- Added a DatasetLoader for loading data from OpenML via scikit-learn.
- Added another R example that solves a regression problem and makes
use of factor variables or categorical variables.
- Added Python 3.7 to Travis CI test settings, removed Python 3.5.
v1.2.0 - 2018-06-29
===================
- You can now run grid searches across a cluster of computers. See
the FAQ entry "Can I use my cluster to run a hyperparameter search?"
for an example.
- PredictService now has better support for text predictions (which
use 1d arrays of strings instead of 2d matrices as inputs) with the
`unwrap_sample` option.
v1.2b1 - 2018-05-24
===================
- Update use of rpy2, no longer used deprecated pandas.rpy module.
- Save `model.cv_results_` in model's metadata if it exists.
- The best model of a grid search can now be persisted using the
`--persist-best` flag.
- Grid search results can now be saved using the `--save-results`
option.
- A `/activate` endpoint allows for activating a certain model version
through the web.
- A `/refit` API endpoint that allows fitting a model to be triggered
through the web.
- A new `/list` API endpoint that works similar to the pld-list CLI
- A `scoring` variable in the configuration root, if defined, will be
used for both pld-test and pld-grid-search
- In CachedUpdatePersister don't load model if we already have active
- Allow for use of multiple CachedUpdatePersisters
Instances of CachedUpdatePersister now use their own name inside the
configuration file to index into `process_store`. That is, a
CachedUpdatePersister that's configured as `my_model_persister`
inside the configuration file will have an entry in `process_store`
with the key `my_model_persister`. This is a breaking change, since
the previously used key `model` is no longer available.
- Added a model persister for REST backends, e.g. for use with
Artifactory.
- Added a `__copy__` directive to copy and override existing entries
in configuration files. This can be used to avoid redundancy in one
or multiple configuration files.
v1.1.0.1 - 2017-04-01
=====================
- Added MANIFEST.in to include VERSION file to distribution
v1.1.0 - 2017-03-27
===================
- Added multiple confuration file support, i.e., PALLADIUM_CONFIG can
now point to a list of configuration files where the configuration
dict is updated successively
- Added functionality to handle multiple predict service entry points
- Added support for nested lists in configuration
- Added `here` variable to be used in configuration pointing to the
parent path of the current configuration file
- Added support for PluggableDecorators to DatasetLoaders
('load_data_decorators' in config)
- Added module to be used with WSGI servers initializing configuration
without delay
- Added parameter for poolclass in sqlalchemy create_engine call in
Database model persister. Default poolclass is set to None now to
avoid DB connections being open for longer time periods
- Updated requirements in order to use newer versions of dependencies
- Added Python 3.6 to Travis CI test settings
- Changed deprecated sklearn.grid_search.GridSearchCV to current
sklearn.model_selection.GridSearchCV
v1.0.1 - 2016-03-04
===================
- Updated requirements in order to use newer versions of dependencies
- Added information about virtual memory size (vms) to /alive entry
point
- Added Python 3.5 to and removed Python 3.3 from Travis CI test
settings
v1.0 - 2015-07-01
=================
- Added a streaming server (`pld-stream`) which listens to stdin,
processes line by line, and returns predictions.
- Added support for batch processing through the web API. You can now
use a POST request to predict for a number of samples at once.
(Please see section "Upgrading" in the docs for details about
changes to the PredictService API.)
- Added support for selecting active model used for predictions.
Previously, the latest trained model was used. (Please see section
"Upgrading" in the docs for details on how to upgrade model
persisters (database or file) to the latest version.
- Added a new command `pld-admin` to activate and delete models.
- Added a new command `pld-dockerize` to create application specific
docker images for Palladium-based services. Previously, this used
to be handled by a script in `addons/docker/palladium_app_image`.
- Added a ModelPersister for RDBMs that don't have BLOB.
- Documented scripts.
v0.9.1 - 2015-03-30
===================
- Added initial version of Palladium
- Added scripts to fit, grid-search, test, and list models
- Added script to provide model as web service
- Added scripts to create Docker images of Palladium-based services
- Added documentation and tutorial
- Added Iris classification as sample application
- Added examples how to integrate R and Julia models
- Added information on how to setup Palladium with Mesos / Marathon