gsarti commited on
Commit
99067b4
1 Parent(s): 3cf1de5

Create README.md

Browse files
Files changed (1) hide show
  1. README.md +93 -0
README.md ADDED
@@ -0,0 +1,93 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ language:
3
+ - it
4
+ license: apache-2.0
5
+ tags:
6
+ - italian
7
+ - sequence-to-sequence
8
+ - style-transfer
9
+ - formality-style-transfer
10
+ datasets:
11
+ - yahoo/xformal_it
12
+ widget:
13
+ - text: "Questa performance è a dir poco spiacevole."
14
+ - text: "In attesa di un Suo cortese riscontro, Le auguriamo un piacevole proseguimento di giornata."
15
+ - text: "Questa visione mi procura una goduria indescrivibile."
16
+ - text: "qualora ciò possa interessarti, ti pregherei di contattarmi."
17
+ metrics:
18
+ - rouge
19
+ - bertscore
20
+ model-index:
21
+ - name: mt5-base-formal-to-informal
22
+ results:
23
+ - task:
24
+ type: formality-style-transfer
25
+ name: "Formal-to-informal Style Transfer"
26
+ dataset:
27
+ type: xformal_it
28
+ name: "XFORMAL (Italian Subset)"
29
+ metrics:
30
+ - type: rouge1
31
+ value: 0.806
32
+ name: "Avg. Test Rouge1"
33
+ - type: rouge2
34
+ value: 0.685
35
+ name: "Avg. Test Rouge2"
36
+ - type: rougeL
37
+ value: 0.796
38
+ name: "Avg. Test RougeL"
39
+ - type: bertscore
40
+ value: 0.818
41
+ name: "Avg. Test BERTScore"
42
+ args:
43
+ - model_type: "dbmdz/bert-base-italian-xxl-uncased"
44
+ - lang: "it"
45
+ - num_layers: 10
46
+ - rescale_with_baseline: True
47
+ - baseline_path: "bertscore_baseline_ita.tsv"
48
+ co2_eq_emissions:
49
+ emissions: "40g"
50
+ source: "Google Cloud Platform Carbon Footprint"
51
+ training_type: "fine-tuning"
52
+ geographical_location: "Eemshaven, Netherlands, Europe"
53
+ hardware_used: "1 TPU v3-8 VM"
54
+ ---
55
+
56
+ # mT5 Base for Formal-to-informal Style Transfer 🤗
57
+
58
+ This repository contains the checkpoint for the [mT5 Base](https://huggingface.co/google/mt5-base) model fine-tuned on Formal-to-informal style transfer on the Italian subset of the XFORMAL dataset as part of the experiments of the paper [IT5: Large-scale Text-to-text Pretraining for Italian Language Understanding and Generation](https://arxiv.org) by Gabriele Sarti and Malvina Nissim.
59
+
60
+ A comprehensive overview of other released materials is provided in the [gsarti/it5](https://github.com/gsarti/it5) repository. Refer to the paper for additional details concerning the reported scores and the evaluation approach.
61
+
62
+ ## Using the model
63
+
64
+ Model checkpoints are available for usage in Tensorflow, Pytorch and JAX. They can be used directly with pipelines as:
65
+
66
+ ```python
67
+ from transformers import pipelines
68
+
69
+ f2i = pipeline("text2text-generation", model='it5/mt5-base-formal-to-informal')
70
+ f2i("Vi ringrazio infinitamente per vostra disponibilità")
71
+ >>> [{"generated_text": "e grazie per la vostra disponibilità!"}]
72
+ ```
73
+
74
+ or loaded using autoclasses:
75
+
76
+ ```python
77
+ from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
78
+
79
+ tokenizer = AutoTokenizer.from_pretrained("it5/mt5-base-formal-to-informal")
80
+ model = AutoModelForSeq2SeqLM.from_pretrained("it5/mt5-base-formal-to-informal")
81
+ ```
82
+
83
+ If you use this model in your research, please cite our work as:
84
+
85
+ ```bibtex
86
+ @article{sarti-nissim-2022-it5,
87
+ title={IT5: Large-scale Text-to-text Pretraining for Italian Language Understanding and Generation},
88
+ author={Sarti, Gabriele and Nissim, Malvina},
89
+ journal={ArXiv preprint TBD},
90
+ url={TBD},
91
+ year={2022}
92
+ }
93
+ ```