claudiatang commited on
Commit
5a0c9be
1 Parent(s): 09374b7

End of training

Browse files
Files changed (2) hide show
  1. README.md +30 -54
  2. pytorch_model.bin +1 -1
README.md CHANGED
@@ -6,42 +6,32 @@ tags:
6
  metrics:
7
  - bleu
8
  model-index:
9
- - name: flan-t5-base-eng-hwp-kjv
10
  results: []
11
- language:
12
- - en
13
- library_name: transformers
14
- pipeline_tag: translation
15
  ---
16
 
17
- # English to Hawaiian Pidgin Translation | flan-t5-base-eng-hwp
 
18
 
19
- This model is a fine-tuned version of [google/flan-t5-base](https://huggingface.co/google/flan-t5-base) on a English and Hawaiian Pidgin dataset.
 
 
20
  It achieves the following results on the evaluation set:
21
- - Loss: 1.5821
22
- - Bleu: 5.0891
23
- - Gen Len: 18.8633
24
 
25
  ## Model description
26
 
27
- ### Running the model
28
-
29
- The [google/flan-t5-base](https://huggingface.co/google/flan-t5-base) documentation has more details on running the model.
30
-
31
- However, to use this model to translate English to Hawaiian Pidgin, enter ``"translate English to Hawaiian Pidgin: "`` before your statement.
32
 
33
- For example, if you would like to translate "I went to Ala Moana today to go shopping" please tokenize all of the following:
34
- ``translate English to Hawaiian Pidgin: I went to Ala Moana today to go shopping.``
35
 
36
- If you are trying the [English-Hawaiian Pidgin Translator](https://huggingface.co/spaces/claudiatang/english_to_hawaiian-pidgin) space, there is no need for the input prefix, as it is automatically added.
37
 
38
  ## Training and evaluation data
39
 
40
- There are not many English-Hawaiian Pidgin parallel corpora that are easily accessible. A parallel dataset, similar to [bible_para](https://huggingface.co/datasets/bible_para), was compiled by scraping the Hawaiʻi Pidgin Version (HWP) and the King James Version (KJV) from [biblegateway.com](https://www.biblegateway.com/). <!--- For more information, please refer to [this notebook](). -->
41
-
42
- ## Intended uses & limitations
43
-
44
- Due to a limited set of training and evaluation data, this model has many limitations, such as not knowing certain Hawaiian Pidgin phrases or having trouble with longer sentences.
45
 
46
  ## Training procedure
47
 
@@ -60,40 +50,26 @@ The following hyperparameters were used during training:
60
 
61
  | Training Loss | Epoch | Step | Validation Loss | Bleu | Gen Len |
62
  |:-------------:|:-----:|:----:|:---------------:|:------:|:-------:|
63
- | No log | 1.0 | 420 | 1.6077 | 3.732 | 18.8506 |
64
- | 2.1314 | 2.0 | 840 | 1.4572 | 4.2893 | 18.8557 |
65
- | 1.5079 | 3.0 | 1260 | 1.3978 | 4.6504 | 18.8599 |
66
- | 1.2945 | 4.0 | 1680 | 1.3788 | 4.8595 | 18.8641 |
67
- | 1.1387 | 5.0 | 2100 | 1.3841 | 4.907 | 18.8819 |
68
- | 1.0142 | 6.0 | 2520 | 1.3776 | 5.0933 | 18.8743 |
69
- | 1.0142 | 7.0 | 2940 | 1.3912 | 5.1246 | 18.8726 |
70
- | 0.9024 | 8.0 | 3360 | 1.4158 | 5.1468 | 18.8692 |
71
- | 0.8227 | 9.0 | 3780 | 1.4403 | 5.1846 | 18.865 |
72
- | 0.749 | 10.0 | 4200 | 1.4685 | 5.0892 | 18.8844 |
73
- | 0.7012 | 11.0 | 4620 | 1.4997 | 5.1485 | 18.8852 |
74
- | 0.6446 | 12.0 | 5040 | 1.5162 | 5.2782 | 18.8776 |
75
- | 0.6446 | 13.0 | 5460 | 1.5465 | 5.0961 | 18.8743 |
76
- | 0.6063 | 14.0 | 5880 | 1.5588 | 5.0723 | 18.8768 |
77
- | 0.5801 | 15.0 | 6300 | 1.5821 | 5.0891 | 18.8633 |
78
 
79
 
80
  ### Framework versions
81
 
82
  - Transformers 4.34.1
83
- - Pytorch 2.0.1+cu118
84
- - Datasets 2.14.5
85
  - Tokenizers 0.14.1
86
-
87
-
88
- ## Resources
89
- - Christodouloupoulos, C., & Steedman, M. (2014). A massively parallel corpus: the Bible in 100 languages. Language Resources and Evaluation, 49(2), 375–395. https://doi.org/10.1007/s10579-014-9287-y
90
-
91
- - Chung, H. W., Hou, L., Longpre, S., Zoph, B., Tay, Y., Fedus, W., … Wei, J. (2022). _Scaling Instruction-Finetuned Language Models._ doi:10.48550/ARXIV.2210.11416
92
-
93
- - _Hawaii Pidgin_. (2017). Wycliffe. https://www.biblegateway.com/versions/Hawaii-Pidgin-HWP/ (Original work published 2000)
94
-
95
- - _King James Bible_. (2017). BibleGateway.com. https://www.biblegateway.com/versions/king-james-version-kjv-bible/ (Original work published 1769)
96
-
97
- - T5. (n.d.). Huggingface.co. https://huggingface.co/docs/transformers/model_doc/t5
98
-
99
- - Translation. (n.d.). Huggingface.co. Retrieved October 18, 2023, from https://huggingface.co/docs/transformers/tasks/translation
 
6
  metrics:
7
  - bleu
8
  model-index:
9
+ - name: flan-t5-base-eng-hwp
10
  results: []
 
 
 
 
11
  ---
12
 
13
+ <!-- This model card has been generated automatically according to the information the Trainer had access to. You
14
+ should probably proofread and complete it, then remove this comment. -->
15
 
16
+ # flan-t5-base-eng-hwp
17
+
18
+ This model is a fine-tuned version of [google/flan-t5-base](https://huggingface.co/google/flan-t5-base) on the None dataset.
19
  It achieves the following results on the evaluation set:
20
+ - Loss: 1.6073
21
+ - Bleu: 4.9074
22
+ - Gen Len: 18.8338
23
 
24
  ## Model description
25
 
26
+ More information needed
 
 
 
 
27
 
28
+ ## Intended uses & limitations
 
29
 
30
+ More information needed
31
 
32
  ## Training and evaluation data
33
 
34
+ More information needed
 
 
 
 
35
 
36
  ## Training procedure
37
 
 
50
 
51
  | Training Loss | Epoch | Step | Validation Loss | Bleu | Gen Len |
52
  |:-------------:|:-----:|:----:|:---------------:|:------:|:-------:|
53
+ | No log | 1.0 | 420 | 1.6234 | 3.464 | 18.8 |
54
+ | 2.1287 | 2.0 | 840 | 1.4706 | 4.1392 | 18.8084 |
55
+ | 1.5094 | 3.0 | 1260 | 1.4294 | 4.3477 | 18.7924 |
56
+ | 1.2917 | 4.0 | 1680 | 1.4015 | 4.5189 | 18.8068 |
57
+ | 1.1347 | 5.0 | 2100 | 1.3900 | 4.697 | 18.8236 |
58
+ | 1.012 | 6.0 | 2520 | 1.4038 | 4.7522 | 18.8051 |
59
+ | 1.012 | 7.0 | 2940 | 1.4086 | 4.8399 | 18.8177 |
60
+ | 0.901 | 8.0 | 3360 | 1.4453 | 4.8191 | 18.8253 |
61
+ | 0.818 | 9.0 | 3780 | 1.4678 | 4.8245 | 18.8203 |
62
+ | 0.7511 | 10.0 | 4200 | 1.4922 | 4.951 | 18.8574 |
63
+ | 0.693 | 11.0 | 4620 | 1.5186 | 4.9174 | 18.8363 |
64
+ | 0.6462 | 12.0 | 5040 | 1.5487 | 5.0009 | 18.8338 |
65
+ | 0.6462 | 13.0 | 5460 | 1.5651 | 5.021 | 18.8295 |
66
+ | 0.6062 | 14.0 | 5880 | 1.5942 | 4.8801 | 18.8245 |
67
+ | 0.5781 | 15.0 | 6300 | 1.6073 | 4.9074 | 18.8338 |
68
 
69
 
70
  ### Framework versions
71
 
72
  - Transformers 4.34.1
73
+ - Pytorch 2.1.0+cu118
74
+ - Datasets 2.14.6
75
  - Tokenizers 0.14.1
 
 
 
 
 
 
 
 
 
 
 
 
 
 
pytorch_model.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:6a46032b703aa3700d619dcfc610d5198c76f993d5731fc18d17041b3720af18
3
  size 990409330
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a07d94c167f066bf9af177fd56cf7185d5d0f7f5abfcde4009e9e6e8b8c48685
3
  size 990409330