stefan-it commited on
Commit
6c16102
1 Parent(s): 19d3037

readme: add initial version

Browse files

This introduces the initial version of model card.

Files changed (1) hide show
  1. README.md +72 -0
README.md ADDED
@@ -0,0 +1,72 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ language: en
3
+ license: mit
4
+ tags:
5
+ - flair
6
+ - token-classification
7
+ - sequence-tagger-model
8
+ base_model: hmteams/teams-base-historic-multilingual-discriminator
9
+ widget:
10
+ - text: On Wednesday , a public dinner was given by the Conservative Burgesses of
11
+ Leads , to the Conservative members of the Leeds Town Council , in the Music Hall
12
+ , Albion-street , which was very numerously attended .
13
+ ---
14
+
15
+ # Fine-tuned Flair Model on TopRes19th English NER Dataset (HIPE-2022)
16
+
17
+ This Flair model was fine-tuned on the
18
+ [TopRes19th English](https://github.com/hipe-eval/HIPE-2022-data/blob/main/documentation/README-topres19th.md)
19
+ NER Dataset using hmTEAMS as backbone LM.
20
+
21
+ The TopRes19th dataset consists of NE-annotated historical English newspaper articles from 19C.
22
+
23
+ The following NEs were annotated: `BUILDING`, `LOC` and `STREET`.
24
+
25
+ # Results
26
+
27
+ We performed a hyper-parameter search over the following parameters with 5 different seeds per configuration:
28
+
29
+ * Batch Sizes: `[8, 4]`
30
+ * Learning Rates: `[3e-05, 5e-05]`
31
+
32
+ And report micro F1-score on development set:
33
+
34
+ | Configuration | Run 1 | Run 2 | Run 3 | Run 4 | Run 5 | Avg. |
35
+ |-----------------|--------------|--------------|--------------|--------------|--------------|--------------|
36
+ | bs8-e10-lr3e-05 | [0.8089][1] | [0.8137][2] | [0.8083][3] | [0.8145][4] | [0.8082][5] | 81.07 ± 0.28 |
37
+ | bs4-e10-lr3e-05 | [0.8068][6] | [0.8008][7] | [0.8195][8] | [0.8086][9] | [0.8049][10] | 80.81 ± 0.63 |
38
+ | bs8-e10-lr5e-05 | [0.818][11] | [0.795][12] | [0.7992][13] | [0.804][14] | [0.7938][15] | 80.2 ± 0.88 |
39
+ | bs4-e10-lr5e-05 | [0.8109][16] | [0.8114][17] | [0.7951][18] | [0.7901][19] | [0.795][20] | 80.05 ± 0.89 |
40
+
41
+ [1]: https://hf.co/stefan-it/hmbench-topres19th-en-hmteams-bs8-wsFalse-e10-lr3e-05-poolingfirst-layers-1-crfFalse-1
42
+ [2]: https://hf.co/stefan-it/hmbench-topres19th-en-hmteams-bs8-wsFalse-e10-lr3e-05-poolingfirst-layers-1-crfFalse-2
43
+ [3]: https://hf.co/stefan-it/hmbench-topres19th-en-hmteams-bs8-wsFalse-e10-lr3e-05-poolingfirst-layers-1-crfFalse-3
44
+ [4]: https://hf.co/stefan-it/hmbench-topres19th-en-hmteams-bs8-wsFalse-e10-lr3e-05-poolingfirst-layers-1-crfFalse-4
45
+ [5]: https://hf.co/stefan-it/hmbench-topres19th-en-hmteams-bs8-wsFalse-e10-lr3e-05-poolingfirst-layers-1-crfFalse-5
46
+ [6]: https://hf.co/stefan-it/hmbench-topres19th-en-hmteams-bs4-wsFalse-e10-lr3e-05-poolingfirst-layers-1-crfFalse-1
47
+ [7]: https://hf.co/stefan-it/hmbench-topres19th-en-hmteams-bs4-wsFalse-e10-lr3e-05-poolingfirst-layers-1-crfFalse-2
48
+ [8]: https://hf.co/stefan-it/hmbench-topres19th-en-hmteams-bs4-wsFalse-e10-lr3e-05-poolingfirst-layers-1-crfFalse-3
49
+ [9]: https://hf.co/stefan-it/hmbench-topres19th-en-hmteams-bs4-wsFalse-e10-lr3e-05-poolingfirst-layers-1-crfFalse-4
50
+ [10]: https://hf.co/stefan-it/hmbench-topres19th-en-hmteams-bs4-wsFalse-e10-lr3e-05-poolingfirst-layers-1-crfFalse-5
51
+ [11]: https://hf.co/stefan-it/hmbench-topres19th-en-hmteams-bs8-wsFalse-e10-lr5e-05-poolingfirst-layers-1-crfFalse-1
52
+ [12]: https://hf.co/stefan-it/hmbench-topres19th-en-hmteams-bs8-wsFalse-e10-lr5e-05-poolingfirst-layers-1-crfFalse-2
53
+ [13]: https://hf.co/stefan-it/hmbench-topres19th-en-hmteams-bs8-wsFalse-e10-lr5e-05-poolingfirst-layers-1-crfFalse-3
54
+ [14]: https://hf.co/stefan-it/hmbench-topres19th-en-hmteams-bs8-wsFalse-e10-lr5e-05-poolingfirst-layers-1-crfFalse-4
55
+ [15]: https://hf.co/stefan-it/hmbench-topres19th-en-hmteams-bs8-wsFalse-e10-lr5e-05-poolingfirst-layers-1-crfFalse-5
56
+ [16]: https://hf.co/stefan-it/hmbench-topres19th-en-hmteams-bs4-wsFalse-e10-lr5e-05-poolingfirst-layers-1-crfFalse-1
57
+ [17]: https://hf.co/stefan-it/hmbench-topres19th-en-hmteams-bs4-wsFalse-e10-lr5e-05-poolingfirst-layers-1-crfFalse-2
58
+ [18]: https://hf.co/stefan-it/hmbench-topres19th-en-hmteams-bs4-wsFalse-e10-lr5e-05-poolingfirst-layers-1-crfFalse-3
59
+ [19]: https://hf.co/stefan-it/hmbench-topres19th-en-hmteams-bs4-wsFalse-e10-lr5e-05-poolingfirst-layers-1-crfFalse-4
60
+ [20]: https://hf.co/stefan-it/hmbench-topres19th-en-hmteams-bs4-wsFalse-e10-lr5e-05-poolingfirst-layers-1-crfFalse-5
61
+
62
+ The [training log](training.log) and TensorBoard logs (only for hmByT5 and hmTEAMS based models) are also uploaded to the model hub.
63
+
64
+ More information about fine-tuning can be found [here](https://github.com/stefan-it/hmBench).
65
+
66
+ # Acknowledgements
67
+
68
+ We thank [Luisa März](https://github.com/LuisaMaerz), [Katharina Schmid](https://github.com/schmika) and
69
+ [Erion Çano](https://github.com/erionc) for their fruitful discussions about Historic Language Models.
70
+
71
+ Research supported with Cloud TPUs from Google's [TPU Research Cloud](https://sites.research.google/trc/about/) (TRC).
72
+ Many Thanks for providing access to the TPUs ❤️