Update config and readme
Browse files- README.md +64 -0
- config.json +10 -10
README.md
CHANGED
|
@@ -1,3 +1,67 @@
|
|
| 1 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 2 |
license: cc-by-4.0
|
| 3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
---
|
| 2 |
+
language: ti
|
| 3 |
+
widget:
|
| 4 |
+
- text: "ድምጻዊ ኣብርሃም ኣፈወርቂ ንዘልኣለም ህያው ኮይኑ ኣብ ልብና ይነብር"
|
| 5 |
+
- text: "ወአመ ሳብዕት ዕለት ቦዘወፅአ እምውስተ ሕዝብ ከመ ያስተጋብእ ወኢረከበ።"
|
| 6 |
+
- text: "እሊ እግል ኖሱ አሳስ ተጠውር ወዐቦት ክምሰልቱ ሸክ ኢወትውዴ።"
|
| 7 |
+
- text: "ኣኩኽር ፡ ልሽክክ ናው ጀረቢነዅስክ ክሙኑኽር ክራውል ሕበርሲድኖ ገረሰነኵ።"
|
| 8 |
+
- text: "ነገ ለግማሽ ፍፃሜ ያለፉትን አሳውቀንና አስመርጠናችሁ እንሸልማለን።"
|
| 9 |
+
tags:
|
| 10 |
+
- geezlab
|
| 11 |
+
metrics:
|
| 12 |
+
- accuracy
|
| 13 |
+
- f1
|
| 14 |
+
- precision
|
| 15 |
+
- recall
|
| 16 |
+
model-index:
|
| 17 |
+
- name: geezswitch-tielectra
|
| 18 |
+
results: []
|
| 19 |
license: cc-by-4.0
|
| 20 |
---
|
| 21 |
+
|
| 22 |
+
# TiELECTRA-GeezSwitch
|
| 23 |
+
|
| 24 |
+
This model is a fine-tuned version of [fgaim/tielectra-small](https://huggingface.co/fgaim/tielectra-small) on the [GeezSwitch](https://github.com/fgaim/geezswitch-data) dataset.
|
| 25 |
+
|
| 26 |
+
It achieves the following results on the test set:
|
| 27 |
+
|
| 28 |
+
- F1: 0.9844
|
| 29 |
+
- Recall: 0.9844
|
| 30 |
+
- Precision: 0.9845
|
| 31 |
+
- Accuracy: 0.9844
|
| 32 |
+
- Loss: 0.2190
|
| 33 |
+
|
| 34 |
+
## Training
|
| 35 |
+
|
| 36 |
+
### Hyperparameters
|
| 37 |
+
|
| 38 |
+
The following hyperparameters were used during training:
|
| 39 |
+
|
| 40 |
+
- learning_rate: 2e-05
|
| 41 |
+
- train_batch_size: 32
|
| 42 |
+
- eval_batch_size: 32
|
| 43 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 44 |
+
- lr_scheduler_type: linear
|
| 45 |
+
- num_epochs: 3.0
|
| 46 |
+
- seed: 42
|
| 47 |
+
|
| 48 |
+
### Framework versions
|
| 49 |
+
|
| 50 |
+
- Transformers 4.19.0.dev0
|
| 51 |
+
- Pytorch 1.11.0+cu113
|
| 52 |
+
- Datasets 2.1.0
|
| 53 |
+
- Tokenizers 0.12.1
|
| 54 |
+
|
| 55 |
+
|
| 56 |
+
### Citation
|
| 57 |
+
|
| 58 |
+
If you use this model or the GeezSwitch model in your research, please cite as follows:
|
| 59 |
+
|
| 60 |
+
```markdown
|
| 61 |
+
@inproceedings{fgaim2022geezswitch,
|
| 62 |
+
title={GeezSwitch: Language Identification in Typologically Related Low-resourced East African Languages},
|
| 63 |
+
author={Fitsum Gaim and Wonsuk Yang and Jong C. Park},
|
| 64 |
+
booktitle={Proceedings of the 13th Language Resources and Evaluation Conference},
|
| 65 |
+
year={2022}
|
| 66 |
+
}
|
| 67 |
+
```
|
config.json
CHANGED
|
@@ -10,20 +10,20 @@
|
|
| 10 |
"hidden_dropout_prob": 0.1,
|
| 11 |
"hidden_size": 256,
|
| 12 |
"id2label": {
|
| 13 |
-
"0": "
|
| 14 |
-
"1": "
|
| 15 |
-
"2": "
|
| 16 |
-
"3": "
|
| 17 |
-
"4": "
|
| 18 |
},
|
| 19 |
"initializer_range": 0.02,
|
| 20 |
"intermediate_size": 1024,
|
| 21 |
"label2id": {
|
| 22 |
-
"
|
| 23 |
-
"
|
| 24 |
-
"
|
| 25 |
-
"
|
| 26 |
-
"
|
| 27 |
},
|
| 28 |
"layer_norm_eps": 1e-12,
|
| 29 |
"max_position_embeddings": 512,
|
|
|
|
| 10 |
"hidden_dropout_prob": 0.1,
|
| 11 |
"hidden_size": 256,
|
| 12 |
"id2label": {
|
| 13 |
+
"0": "Amharic",
|
| 14 |
+
"1": "Blin",
|
| 15 |
+
"2": "Ge'ez",
|
| 16 |
+
"3": "Tigre",
|
| 17 |
+
"4": "Tigrinya"
|
| 18 |
},
|
| 19 |
"initializer_range": 0.02,
|
| 20 |
"intermediate_size": 1024,
|
| 21 |
"label2id": {
|
| 22 |
+
"Amharic": 0,
|
| 23 |
+
"Blin": 1,
|
| 24 |
+
"Ge'ez": 2,
|
| 25 |
+
"Tigre": 3,
|
| 26 |
+
"Tigrinya": 4
|
| 27 |
},
|
| 28 |
"layer_norm_eps": 1e-12,
|
| 29 |
"max_position_embeddings": 512,
|