-
Notifications
You must be signed in to change notification settings - Fork 2
/
CHANGE.txt
82 lines (63 loc) · 2.62 KB
/
CHANGE.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
1.4.1:
1. [feat] remove helper for dl, mxnet and torch
2. [refactor] lib is renamed to utils
3. [refactor] arch is renamed to infrastructure
1.3.37:
1. [feat] add workspace
1.3.36:
1. [chore] replace `multiprocessing` with `multiprocess`
1.3.35:
1. [bugfix] gpu device specification in `set_device`
2. [feat] can use dict-type select in torch trainer
3. [feat] enable milliseconds display in light module
1.3.34:
1. [bugfix] missing `()` in `set_device` of `PytorchHelper`
1.3.33:
1. replace `max_rows` with `display.max_rows` and `max_columns` with `display.max_columns` to support pandas 1.4
1.3.32:
1. enable different ctx for train and eval in light_module
2. fix some errors in light_module
3. add verbose keyword in light_module
1.3.31:
1. add instance_var in Configuration class
2. fix invalid hyper_search with nni in light_module
3. add verbose and dump_cfg keywords in light_module
4. add some new functions: block_std, get_params_filepath, params_filename
1.3.30:
1. add update method in Configuration class
1.3.29:
1. mxnet gallery is moved to Baize(https://github.com/tswsxk/Baize)
2. update light_module
1.3.28:
1. resolve cross import errors in PytorchHelper
1.3.27:
1. fix several errors in PytorchHelper
1.3.26:
1. remove auc in ranking_report when ground_truth is set as continuous
2. fix some errors caused by data parallel of pytorch, including save_params and set_device in PytorchHelper
3. add some general used regex select in ML
4. Configuration class support get method
1.3.25:
1. RandomState is replaced by default_rng in ML PairSampler and add fast implicit sampling in ML PairSampler
2. Update arguments and default actions in ranking_report
3. Add parent_dir method in lib.path
v1.3.24:
1. change dependencies, remove matplotlib, tensorboard, tensorboardx in ml;
2. new installation mode: viz with matplotlib, tensorboard, tensorboardx, ml-viz with basic dependencies plus viz deps.
v1.3.23:
1. remove EvalFormatter, which is replaced by EpochEvalFMT
2. remove tmt_value and tmt_loss function
1.3.22:
1. remove InheritableConfiguration class in lib/parser
2. add ranking metrics
3. add sampler in ML packages
1.3.21:
1. Remove lib/structure/Factor.py
1.3.20:
1. Remove deprecated methods ResultAnalyser.
2. fix some style errors.
3. Add light module for pytorch
1.3.19:
1. Update optimizer_cfg.py in DL package: update the default learning rate and weight decay
2. Change fit function in DL package: reassign the args locations
3. Add light module (unstable)