Pengcheng He commited on
Commit
1f2d52b
1 Parent(s): 1a06341

Create deberta xlarge 750M model

Browse files
Files changed (5) hide show
  1. README.md +38 -0
  2. bpe_encoder.bin +3 -0
  3. config.json +18 -0
  4. pytorch_model.bin +3 -0
  5. tokenizer_config.json +3 -0
README.md ADDED
@@ -0,0 +1,38 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ thumbnail: https://huggingface.co/front/thumbnails/microsoft.png
3
+ license: mit
4
+ ---
5
+
6
+ ## DeBERTa: Decoding-enhanced BERT with Disentangled Attention
7
+
8
+ [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.
9
+
10
+ Please check the [official repository](https://github.com/microsoft/DeBERTa) for more details and updates.
11
+
12
+ This the DeBERTa xlarge model with 48 layers, 1024 hidden size. Total parameters 750M.
13
+
14
+ #### Fine-tuning on NLU tasks
15
+
16
+ We present the dev results on SQuAD 1.1/2.0 and several GLUE benchmark tasks.
17
+
18
+ | Model | SQuAD 1.1 | SQuAD 2.0 | MNLI-m | SST-2 | QNLI | CoLA | RTE | MRPC | QQP |STS-B|
19
+ |-------------------|-----------|-----------|--------|-------|------|------|------|------|------|-----|
20
+ | BERT-Large | 90.9/84.1 | 81.8/79.0 | 86.6 | 93.2 | 92.3 | 60.6 | 70.4 | 88.0 | 91.3 |90.0 |
21
+ | RoBERTa-Large | 94.6/88.9 | 89.4/86.5 | 90.2 | 96.4 | 93.9 | 68.0 | 86.6 | 90.9 | 92.2 |92.4 |
22
+ | XLNet-Large | 95.1/89.7 | 90.6/87.9 | 90.8 | 97.0 | 94.9 | 69.0 | 85.9 | 90.8 | 92.3 |92.5 |
23
+ | **DeBERTa-Large** | 95.5/90.1 | 90.7/88.0 | 91.1 | 96.5 | 95.3 | 69.5 | 88.1 | 92.5 | 92.3 |92.5 |
24
+
25
+ ### Citation
26
+
27
+ If you find DeBERTa useful for your work, please cite the following paper:
28
+
29
+ ``` latex
30
+ @misc{he2020deberta,
31
+ title={DeBERTa: Decoding-enhanced BERT with Disentangled Attention},
32
+ author={Pengcheng He and Xiaodong Liu and Jianfeng Gao and Weizhu Chen},
33
+ year={2020},
34
+ eprint={2006.03654},
35
+ archivePrefix={arXiv},
36
+ primaryClass={cs.CL}
37
+ }
38
+ ```
bpe_encoder.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e7c6f9eecb461c01e09c00656ccf3e27944b9e74bfe29e51632b13d3cd9d6c8e
3
+ size 3917897
config.json ADDED
@@ -0,0 +1,18 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "attention_probs_dropout_prob": 0.1,
3
+ "hidden_act": "gelu",
4
+ "hidden_dropout_prob": 0.1,
5
+ "hidden_size": 1024,
6
+ "initializer_range": 0.02,
7
+ "intermediate_size": 4096,
8
+ "max_position_embeddings": 512,
9
+ "relative_attention": true,
10
+ "pos_att_type": "c2p|p2c",
11
+ "layer_norm_eps": 1e-7,
12
+ "max_relative_positions": -1,
13
+ "position_biased_input": false,
14
+ "num_attention_heads": 16,
15
+ "num_hidden_layers": 48,
16
+ "type_vocab_size": 0,
17
+ "vocab_size": 50265
18
+ }
pytorch_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:34618faca5332b6ebf078c8e243c0f426c3dc39e299821d492530ef16ae81477
3
+ size 1518990915
tokenizer_config.json ADDED
@@ -0,0 +1,3 @@
 
 
 
1
+ {
2
+ "do_lower_case": false
3
+ }