-
Notifications
You must be signed in to change notification settings - Fork 7
/
CHANGELOG.txt
146 lines (120 loc) · 6.36 KB
/
CHANGELOG.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
=====================
TextWiser CHANGELOG
=====================
-------------------------------------------------------------------------------
Sep 9, 2024 2.0.1
-------------------------------------------------------------------------------
minor:
- Added `bpemb` to requirements as it is no longer a mandatory requirement for flair
-------------------------------------------------------------------------------
May 30, 2024 2.0.0
-------------------------------------------------------------------------------
major:
- Update CI test environment to Python 3.8, 3.9, 3.10, 3.11 and drop 3.7
- Remove support for Python 3.6 and Python 3.7
- Fix scipy version temporarily until `gensim` releases a patch for the following issue
Source: https://github.com/piskvorky/gensim/issues/3525
- Temporarily commented three tests in `test_doc2vec.py` for the same reason above
Source: https://github.com/fidelity/textwiser/blob/master/tests/test_doc2vec.py
-------------------------------------------------------------------------------
Feb 13, 2024 1.5.1
-------------------------------------------------------------------------------
minor:
- Change the binary parameter of sklearn models to be a boolean
Source: https://github.com/fidelity/textwiser/blob/2eb5d19c82f357e6d8e3fce5f4aa65bf71312100/textwiser/embeddings/random.py#L41
- Fix a test issue caused by umap fit_transform
The issue is caused by the new scipy version. So, issue is fixed by commenting the assertion and adding the version with which it would pass
Source: https://github.com/fidelity/textwiser/blob/fix/failed_tests/tests/test_umap.py#L23
-------------------------------------------------------------------------------
Dec 19, 2022 1.5.0
-------------------------------------------------------------------------------
major:
- Utilize ELMo from TFHub and remove allennlp dependency
-------------------------------------------------------------------------------
Mar 03, 2022 1.4.0
-------------------------------------------------------------------------------
major:
- Update UMAP to return deterministic output in latest version
Source: https://github.com/fidelity/textwiser/blob/39bad042104c41d0d57174b49941882af79cc3db/textwiser/transformations/umap_.py#L23
- Update SVD to return deterministic output
Source: https://github.com/fidelity/textwiser/blob/39bad042104c41d0d57174b49941882af79cc3db/textwiser/transformations/svd.py#L26
- Update Word2Vec and Doc2Vec with latest gensim training
Source: https://github.com/fidelity/textwiser/blob/39bad042104c41d0d57174b49941882af79cc3db/textwiser/embeddings/word.py#L204
Source: https://github.com/fidelity/textwiser/blob/39bad042104c41d0d57174b49941882af79cc3db/textwiser/embeddings/doc2vec.py#L25
minor:
- Disable non-relevant transformers warning
- Add full requirement install option (see requirements_full.txt)
- Make gensim requirement explicit instead of relying on flair's dependency tree
- Directly utilize ELMo from allennlp instead of flair by bumping allennlp requirement
-------------------------------------------------------------------------------
Feb 23, 2022 1.3.2
-------------------------------------------------------------------------------
minor:
- Added upper bound for flair (less or equal to 0.9) to pass python3.9 unit tests
-------------------------------------------------------------------------------
Jan 21, 2021 1.3.1
-------------------------------------------------------------------------------
minor:
- Updated dependency requirements to reflect updates to Flair library
-------------------------------------------------------------------------------
Jul 22, 2020 1.3.0
-------------------------------------------------------------------------------
major:
- Added T5, XLM-RoBERTa, BART, ECELTRA, DialoGPT, Longformer embeddings
- Added inline pooling for Word embeddings to make some pooling scenarios faster
minor:
- Updated basic usage notebook
- Updated preferred method of persistence & examples
- SVD now properly appears in model parameters and can be fine-tuned
-------------------------------------------------------------------------------
Mar 12, 2020 1.2.1
-------------------------------------------------------------------------------
minor:
- Fix transformers model initialization for newer versions of transformers
-------------------------------------------------------------------------------
Feb 13, 2020 1.2.0
-------------------------------------------------------------------------------
major:
- Update Universal Sentence Encoder
- Now works with newer versions
- Default version is now v5
- Add bytepair embeddings
- Add ALBERT embeddings
minor:
- Add explicit fit_transform method to Compound embeddings to make them faster
-------------------------------------------------------------------------------
Nov 13, 2019 1.1.1
-------------------------------------------------------------------------------
minor:
- Fix running transformers models on a GPU
- Fix modification of input schema object for Compound embeddings
-------------------------------------------------------------------------------
Oct 24, 2019 1.1.0
-------------------------------------------------------------------------------
major:
- All transformer-based models are powered by the transformers library
- This enables fine-tuning for transformer models!
- There isn't a subword to word level aggregation, everything is aggregated at doc level
- The models output the last layer by default
- The scalar mix parameter is discontinued
- Add DistilBERT and CTRL models
- Add Sphinx documentation
- Add LICENSE
minor:
- More modular word embedding testing
-------------------------------------------------------------------------------
Oct 3, 2019 1.0.1
-------------------------------------------------------------------------------
major:
- Data in different structures are only converted to torch tensor if necessary (major speedup for sklearn models!)
minor:
- Fix word2vec fit behavior to match other trainable models
- Add explicit fit_transform to TextWiser for faster inference with sklearn models
-------------------------------------------------------------------------------
Sept 27, 2019 1.0.0
-------------------------------------------------------------------------------
- Initial release.
-------------------------------------------------------------------------------
July 29, 2019 0.0.1
-------------------------------------------------------------------------------
- Development starts. Unstable.