Commit
•
2a51ca8
1
Parent(s):
3cf7065
Create README.md
Browse files
README.md
ADDED
|
@@ -0,0 +1,50 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
language:
|
| 3 |
+
- ms
|
| 4 |
+
tags:
|
| 5 |
+
- paraphrase
|
| 6 |
+
metrics:
|
| 7 |
+
- sacrebleu
|
| 8 |
+
---
|
| 9 |
+
|
| 10 |
+
# finetune-paraphrase-t5-base-standard-bahasa-cased
|
| 11 |
+
|
| 12 |
+
Finetuned T5 base on MS paraphrase tasks.
|
| 13 |
+
|
| 14 |
+
## Dataset
|
| 15 |
+
|
| 16 |
+
1. translated PAWS, https://huggingface.co/datasets/mesolitica/translated-PAWS
|
| 17 |
+
2. translated MRPC, https://huggingface.co/datasets/mesolitica/translated-MRPC
|
| 18 |
+
|
| 19 |
+
## Finetune details
|
| 20 |
+
|
| 21 |
+
1. Finetune using single RTX 3090 Ti.
|
| 22 |
+
|
| 23 |
+
Scripts at https://github.com/huseinzol05/malaya/tree/master/session/paraphrase/hf-t5
|
| 24 |
+
|
| 25 |
+
## Supported prefix
|
| 26 |
+
|
| 27 |
+
1. `parafrasa: {string}`, for MS paraphrase.
|
| 28 |
+
|
| 29 |
+
## Evaluation
|
| 30 |
+
|
| 31 |
+
Evaluated on MRPC validation set and PAWS test set.
|
| 32 |
+
|
| 33 |
+
```
|
| 34 |
+
{'name': 'BLEU',
|
| 35 |
+
'score': 58.764876478744064,
|
| 36 |
+
'_mean': -1.0,
|
| 37 |
+
'_ci': -1.0,
|
| 38 |
+
'_verbose': '84.5/65.8/53.0/43.1 (BP = 0.984 ratio = 0.985 hyp_len = 138828 ref_len = 141004)',
|
| 39 |
+
'bp': 0.9844481274610702,
|
| 40 |
+
'counts': [117307, 86700, 66087, 50665],
|
| 41 |
+
'totals': [138828, 131754, 124680, 117606],
|
| 42 |
+
'sys_len': 138828,
|
| 43 |
+
'ref_len': 141004,
|
| 44 |
+
'precisions': [84.49808396000807,
|
| 45 |
+
65.80445375472472,
|
| 46 |
+
53.00529355149182,
|
| 47 |
+
43.080285019471795],
|
| 48 |
+
'prec_str': '84.5/65.8/53.0/43.1',
|
| 49 |
+
'ratio': 0.9845678136790446}
|
| 50 |
+
```
|