/
bert-base-multilingual-cased-de-wiki_houlsby.yaml
82 lines (66 loc) 路 2.92 KB
/
bert-base-multilingual-cased-de-wiki_houlsby.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
# Adapter-Hub adapter entry
# Defines a single adapter entry in Adapter-Hub
# --------------------
# The name of the author(s) of this adapter.
author: "Jonas Pfeiffer"
# A bibtex citation of the work related to this adapter.
citation: |
@article{pfeiffer20madx,
title={{MAD-X}: An {A}dapter-based {F}ramework for {M}ulti-task {C}ross-lingual {T}ransfer},
author={Pfeiffer, Jonas and Vuli\'{c}, Ivan and Gurevych, Iryna and Ruder, Sebastian},
journal={arXiv preprint},
year={2020},
url={https://arxiv.org/pdf/2005.00052.pdf},
}
# Describes the adapter architecture used by this adapter
config: # TODO: REQUIRED
# The name of the adapter config used by this adapter (a short name available in the `architectures` folder).
# Example: pfeiffer
using: houlsby
# Overrides the default activation function of the specified adapter architecture.
# Example: tanh
non_linearity: gelu
# Overrides the default reduction factor of the specified adapter architecture
# Example: 64
reduction_factor: 2
# The version to be downloaded if no version is explicitly stated.
default_version: "nd_200k"
# A short description of this adapter.
description: |
Houlsby Adapter trained with Masked Language Modelling on German Wikipedia Articles for 250k steps and a batch size of 64.
# A contact email of the author(s).
email: "pfeiffer@ukp.informatik.tu-darmstadt"
# A list of different versions of this adapter available for download.
files:
- sha1: "ef86eaba212e9210a20bf6aaaf233f5b1c485d4c"
sha256: "c2f57b1cf1da67871779f0777ca50541d19dbaa679279cfb23c67a2f3352d71c"
# Download URL pointing to a zip folder containing the adapter module.
url: "https://public.ukp.informatik.tu-darmstadt.de/AdapterHub/text_lang/de/bert-base-multilingual-cased/houlsby/de_houlsby_gelu_nd_200k.zip"
version: "nd_200k"
- sha1: "8bd54d863ced28961d44f9a584a62042774f495b"
sha256: "4ea06e7032f9019019d1469e6649f5a2e5abf3b2f0cfd7a13036f2080f5c43bd"
# Download URL pointing to a zip folder containing the adapter module.
url: "https://public.ukp.informatik.tu-darmstadt.de/AdapterHub/text_lang/de/bert-base-multilingual-cased/houlsby/de_houlsby_gelu_200k.zip"
version: "wd_200k"
# - ...
# A GitHub handle associated with the author(s).
github: "jopfeiff"
# The hidden size of the model
hidden_size: 768
# The string identifier of the pre-trained model (by which it is identified at Huggingface).
# Example: bert-base-uncased
model_name: "bert-base-multilingual-cased"
# The model type.
# Example: bert
model_type: bert
# The string identifier of the subtask this adapter belongs to.
subtask: wiki
# The string identifier of the task this adapter belongs to.
task: de
# A Twitter handle associated with the author(s).
twitter: "@PfeiffJo"
# The type of adapter (one of the options available in `adapter_type`.
type: "text_lang"
# A URL providing more information on this adapter/ the authors/ the organization.
url: "https://pfeiffer.ai"
prediction_head: false