Hugging Face's logo

stefan-it
/
hmbench-hipe2020-fr-hmteams-bs8-wsFalse-e10-lr3e-05-poolingfirst-layers-1-crfFalse-2

Token Classification
Flair
PyTorch
TensorBoard
French
sequence-tagger-model
Model card Files Metrics Community
1
stefan-it commited on
Commit
445d5da
•
1 Parent(s): e849875

readme: add initial version of model card

Browse files

Hi,

this PR add the initial version of model card.

Files changed (1) hide show
  1. README.md +75 -0
README.md ADDED
@@ -0,0 +1,75 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ language: fr
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: Nous recevons le premier numéro d ' un nouveau journal , le Radical - Libéral
11
+ , qui paraîtra à Genève deux fois la semaine . Son but est de représenter l '
12
+ élément national du radicalisme genevois , en d ' autres termes , de défendre
13
+ la politique intransigeante do M . Carteret , en opposition aux tendances du groupe
14
+ _ > dont le Genevois est l ' organe . Bétail .
15
+ ---
16
+
17
+ # Fine-tuned Flair Model on French HIPE-2020 Dataset (HIPE-2022)
18
+
19
+ This Flair model was fine-tuned on the
20
+ [French HIPE-2020](https://github.com/hipe-eval/HIPE-2022-data/blob/main/documentation/README-hipe2020.md)
21
+ NER Dataset using hmTEAMS as backbone LM.
22
+
23
+ The HIPE-2020 dataset is comprised of newspapers from mid 19C to mid 20C. For information can be found
24
+ [here](https://dl.acm.org/doi/abs/10.1007/978-3-030-58219-7_21).
25
+
26
+ The following NEs were annotated: `loc`, `org`, `pers`, `prod`, `time` and `comp`.
27
+
28
+ # Results
29
+
30
+ We performed a hyper-parameter search over the following parameters with 5 different seeds per configuration:
31
+
32
+ * Batch Sizes: `[8, 4]`
33
+ * Learning Rates: `[3e-05, 5e-05]`
34
+
35
+ And report micro F1-score on development set:
36
+
37
+ | Configuration | Run 1 | Run 2 | Run 3 | Run 4 | Run 5 | Avg. |
38
+ |-----------------|-------------|--------------|--------------|--------------|--------------|--------------|
39
+ | bs8-e10-lr3e-05 | [0.8652][1] | [0.8552][2] | [0.8526][3] | [0.8585][4] | [0.8615][5] | 85.86 ± 0.45 |
40
+ | bs4-e10-lr3e-05 | [0.8602][6] | [0.8556][7] | [0.8512][8] | [0.8511][9] | [0.8595][10] | 85.55 ± 0.39 |
41
+ | bs8-e10-lr5e-05 | [0.85][11] | [0.8592][12] | [0.8534][13] | [0.8473][14] | [0.8505][15] | 85.21 ± 0.41 |
42
+ | bs4-e10-lr5e-05 | [0.839][16] | [0.8473][17] | [0.8385][18] | [0.839][19] | [0.8488][20] | 84.25 ± 0.45 |
43
+
44
+ [1]: https://hf.co/stefan-it/hmbench-hipe2020-fr-hmteams-bs8-wsFalse-e10-lr3e-05-poolingfirst-layers-1-crfFalse-1
45
+ [2]: https://hf.co/stefan-it/hmbench-hipe2020-fr-hmteams-bs8-wsFalse-e10-lr3e-05-poolingfirst-layers-1-crfFalse-2
46
+ [3]: https://hf.co/stefan-it/hmbench-hipe2020-fr-hmteams-bs8-wsFalse-e10-lr3e-05-poolingfirst-layers-1-crfFalse-3
47
+ [4]: https://hf.co/stefan-it/hmbench-hipe2020-fr-hmteams-bs8-wsFalse-e10-lr3e-05-poolingfirst-layers-1-crfFalse-4
48
+ [5]: https://hf.co/stefan-it/hmbench-hipe2020-fr-hmteams-bs8-wsFalse-e10-lr3e-05-poolingfirst-layers-1-crfFalse-5
49
+ [6]: https://hf.co/stefan-it/hmbench-hipe2020-fr-hmteams-bs4-wsFalse-e10-lr3e-05-poolingfirst-layers-1-crfFalse-1
50
+ [7]: https://hf.co/stefan-it/hmbench-hipe2020-fr-hmteams-bs4-wsFalse-e10-lr3e-05-poolingfirst-layers-1-crfFalse-2
51
+ [8]: https://hf.co/stefan-it/hmbench-hipe2020-fr-hmteams-bs4-wsFalse-e10-lr3e-05-poolingfirst-layers-1-crfFalse-3
52
+ [9]: https://hf.co/stefan-it/hmbench-hipe2020-fr-hmteams-bs4-wsFalse-e10-lr3e-05-poolingfirst-layers-1-crfFalse-4
53
+ [10]: https://hf.co/stefan-it/hmbench-hipe2020-fr-hmteams-bs4-wsFalse-e10-lr3e-05-poolingfirst-layers-1-crfFalse-5
54
+ [11]: https://hf.co/stefan-it/hmbench-hipe2020-fr-hmteams-bs8-wsFalse-e10-lr5e-05-poolingfirst-layers-1-crfFalse-1
55
+ [12]: https://hf.co/stefan-it/hmbench-hipe2020-fr-hmteams-bs8-wsFalse-e10-lr5e-05-poolingfirst-layers-1-crfFalse-2
56
+ [13]: https://hf.co/stefan-it/hmbench-hipe2020-fr-hmteams-bs8-wsFalse-e10-lr5e-05-poolingfirst-layers-1-crfFalse-3
57
+ [14]: https://hf.co/stefan-it/hmbench-hipe2020-fr-hmteams-bs8-wsFalse-e10-lr5e-05-poolingfirst-layers-1-crfFalse-4
58
+ [15]: https://hf.co/stefan-it/hmbench-hipe2020-fr-hmteams-bs8-wsFalse-e10-lr5e-05-poolingfirst-layers-1-crfFalse-5
59
+ [16]: https://hf.co/stefan-it/hmbench-hipe2020-fr-hmteams-bs4-wsFalse-e10-lr5e-05-poolingfirst-layers-1-crfFalse-1
60
+ [17]: https://hf.co/stefan-it/hmbench-hipe2020-fr-hmteams-bs4-wsFalse-e10-lr5e-05-poolingfirst-layers-1-crfFalse-2
61
+ [18]: https://hf.co/stefan-it/hmbench-hipe2020-fr-hmteams-bs4-wsFalse-e10-lr5e-05-poolingfirst-layers-1-crfFalse-3
62
+ [19]: https://hf.co/stefan-it/hmbench-hipe2020-fr-hmteams-bs4-wsFalse-e10-lr5e-05-poolingfirst-layers-1-crfFalse-4
63
+ [20]: https://hf.co/stefan-it/hmbench-hipe2020-fr-hmteams-bs4-wsFalse-e10-lr5e-05-poolingfirst-layers-1-crfFalse-5
64
+
65
+ The [training log](training.log) and TensorBoard logs (only for hmByT5 and hmTEAMS based models) are also uploaded to the model hub.
66
+
67
+ More information about fine-tuning can be found [here](https://github.com/stefan-it/hmBench).
68
+
69
+ # Acknowledgements
70
+
71
+ We thank [Luisa März](https://github.com/LuisaMaerz), [Katharina Schmid](https://github.com/schmika) and
72
+ [Erion Çano](https://github.com/erionc) for their fruitful discussions about Historic Language Models.
73
+
74
+ Research supported with Cloud TPUs from Google's [TPU Research Cloud](https://sites.research.google/trc/about/) (TRC).
75
+ Many Thanks for providing access to the TPUs ❤️