Pengcheng He commited on
Commit
5f32929
1 Parent(s): 8dee314

Add deberta v3 base model

Browse files
Files changed (5) hide show
  1. README.md +45 -0
  2. config.json +22 -0
  3. pytorch_model.bin +3 -0
  4. spm.model +3 -0
  5. tokenizer_config.json +4 -0
README.md ADDED
@@ -0,0 +1,45 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ language: en
3
+ tags:
4
+ - deberta
5
+ - deberta-v3
6
+ thumbnail: https://huggingface.co/front/thumbnails/microsoft.png
7
+ license: mit
8
+ ---
9
+
10
+ ## DeBERTa: Decoding-enhanced BERT with Disentangled Attention
11
+
12
+ [DeBERTa](https://arxiv.org/abs/2006.03654) improves the BERT and RoBERTa models using disentangled attention and enhanced mask decoder. With those two improvements, DeBERTa out perform RoBERTa on a majority of NLU tasks with 80GB training data.
13
+
14
+ Please check the [official repository](https://github.com/microsoft/DeBERTa) for more details and updates.
15
+
16
+ In DeBERTa V3 we replaced MLM objective with RTD(Replaced Token Detection) objective during pre-training, which significantly improves the model performance. Please check appendix A11 in our paper DeBERTa for more details.
17
+
18
+ This is the DeBERTa V3 small model with 12 layers, 768 hidden size. Total parameters is 183M while Embedding layer take about 98M due to the usage of 128k vocabulary. It's trained with 160GB data.
19
+ #### Fine-tuning on NLU tasks
20
+
21
+ We present the dev results on SQuAD 1.1/2.0 and MNLI tasks.
22
+
23
+ | Model | SQuAD 1.1 | SQuAD 2.0 | MNLI-m |
24
+ |-------------------|-----------|-----------|--------|
25
+ | RoBERTa-base | 91.5/84.6 | 83.7/80.5 | 87.6 |
26
+ | XLNet-base | -/- | -/80.2 | 86.8 |
27
+ | DeBERTa-base | 93.1/87.2 | 86.2/83.1 | 88.8 |
28
+ | **DeBERTa-v3-base** | 93.9/88.4 | 88.4/85.4 | 90.6 |
29
+ | DeBERTa-v3-base+SiFT | -/- | -/- | **91.0** |
30
+
31
+
32
+ ### Citation
33
+
34
+ If you find DeBERTa useful for your work, please cite the following paper:
35
+
36
+ ``` latex
37
+ @inproceedings{
38
+ he2021deberta,
39
+ title={DEBERTA: DECODING-ENHANCED BERT WITH DISENTANGLED ATTENTION},
40
+ author={Pengcheng He and Xiaodong Liu and Jianfeng Gao and Weizhu Chen},
41
+ booktitle={International Conference on Learning Representations},
42
+ year={2021},
43
+ url={https://openreview.net/forum?id=XPZIaotutsD}
44
+ }
45
+ ```
config.json ADDED
@@ -0,0 +1,22 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model_type": "deberta-v2",
3
+ "attention_probs_dropout_prob": 0.1,
4
+ "hidden_act": "gelu",
5
+ "hidden_dropout_prob": 0.1,
6
+ "hidden_size": 768,
7
+ "initializer_range": 0.02,
8
+ "intermediate_size": 3072,
9
+ "max_position_embeddings": 512,
10
+ "relative_attention": true,
11
+ "position_buckets": 256,
12
+ "norm_rel_ebd": "layer_norm",
13
+ "share_att_key": true,
14
+ "pos_att_type": "p2c|c2p",
15
+ "layer_norm_eps": 1e-7,
16
+ "max_relative_positions": -1,
17
+ "position_biased_input": false,
18
+ "num_attention_heads": 12,
19
+ "num_hidden_layers": 12,
20
+ "type_vocab_size": 0,
21
+ "vocab_size": 128100
22
+ }
pytorch_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:31334fde3063dc2c07242d8633d0aeb0402771f67eaa9ee2eb61fe71e962d99c
3
+ size 372331697
spm.model ADDED
@@ -0,0 +1,3 @@
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c679fbf93643d19aab7ee10c0b99e460bdbc02fedf34b92b05af343b4af586fd
3
+ size 2464616
tokenizer_config.json ADDED
@@ -0,0 +1,4 @@
 
 
 
 
1
+ {
2
+ "do_lower_case": false,
3
+ "vocab_type": "spm"
4
+ }