julien-c commited on
Commit
1acad16
1 Parent(s): 2d8f4d8

Migrate model card from transformers-repo

Browse files

Read announcement at https://discuss.huggingface.co/t/announcement-all-model-cards-will-be-migrated-to-hf-co-model-repos/2755
Original file history: https://github.com/huggingface/transformers/commits/master/model_cards/microsoft/xprophetnet-large-wiki100-cased/README.md

Files changed (1) hide show
  1. README.md +42 -0
README.md ADDED
@@ -0,0 +1,42 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ language: multilingual
3
+ ---
4
+
5
+ ## xprophetnet-large-wiki100-cased
6
+ Cross-lingual version [ProphetNet](https://arxiv.org/abs/2001.04063), pretrained on [wiki100 xGLUE dataset](https://arxiv.org/abs/2004.01401).
7
+ ProphetNet is a new pre-trained language model for sequence-to-sequence learning with a novel self-supervised objective called future n-gram prediction.
8
+ ProphetNet is able to predict more future tokens with a n-stream decoder. The original implementation is Fairseq version at [github repo](https://github.com/microsoft/ProphetNet).
9
+
10
+ xProphetNet is also served as the baseline model for xGLUE cross-lingual natural language generation tasks.
11
+ For xGLUE corss-lingual NLG tasks, xProphetNet is finetuned with English data, but inference with both English and other zero-shot language data.
12
+
13
+ ### Usage
14
+
15
+ This pre-trained model can be fine-tuned on *sequence-to-sequence* tasks. The model could *e.g.* be trained on English headline generation as follows:
16
+
17
+ ```python
18
+ from transformers import XLMProphetNetForConditionalGeneration, XLMProphetNetTokenizer
19
+
20
+ model = XLMProphetNetForConditionalGeneration.from_pretrained("microsoft/xprophetnet-large-wiki100-cased")
21
+ tokenizer = XLMProphetNetTokenizer.from_pretrained("microsoft/xprophetnet-large-wiki100-cased")
22
+
23
+ input_str = "the us state department said wednesday it had received no formal word from bolivia that it was expelling the us ambassador there but said the charges made against him are `` baseless ."
24
+ target_str = "us rejects charges against its ambassador in bolivia"
25
+
26
+ input_ids = tokenizer(input_str, return_tensors="pt").input_ids
27
+ labels = tokenizer(target_str, return_tensors="pt").input_ids
28
+
29
+ loss = model(input_ids, labels=labels).loss
30
+ ```
31
+
32
+ Note that since this model is a multi-lingual model it can be fine-tuned on all kinds of other languages.
33
+
34
+ ### Citation
35
+ ```bibtex
36
+ @article{yan2020prophetnet,
37
+ title={Prophetnet: Predicting future n-gram for sequence-to-sequence pre-training},
38
+ author={Yan, Yu and Qi, Weizhen and Gong, Yeyun and Liu, Dayiheng and Duan, Nan and Chen, Jiusheng and Zhang, Ruofei and Zhou, Ming},
39
+ journal={arXiv preprint arXiv:2001.04063},
40
+ year={2020}
41
+ }
42
+ ```