Pengcheng He commited on
Commit
bbc9e5a
1 Parent(s): 3ecb335

DeBERTa v2 XXLarge(1.5B) model

Browse files
Files changed (5) hide show
  1. README.md +55 -0
  2. config.json +24 -0
  3. pytorch_model.bin +3 -0
  4. spm.model +3 -0
  5. tokenizer_config.json +4 -0
README.md ADDED
@@ -0,0 +1,55 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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 is the DeBERTa V2 xxlarge model with 48 layers, 1536 hidden size. Total parameters 1.5B. It's trained with 160GB data.
13
+
14
+
15
+ ### Fine-tuning on NLU tasks
16
+
17
+ We present the dev results on SQuAD 1.1/2.0 and several GLUE benchmark tasks.
18
+
19
+ | Model | SQuAD 1.1 | SQuAD 2.0 | MNLI-m/mm | SST-2 | QNLI | CoLA | RTE | MRPC | QQP |STS-B |
20
+ |---------------------------|-----------|-----------|-------------|-------|------|------|--------|-------|-------|------|
21
+ | | F1/EM | F1/EM | Acc | Acc | Acc | MCC | Acc |Acc/F1 |Acc/F1 |P/S |
22
+ | 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/- |
23
+ | 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/- |
24
+ | 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/- |
25
+ | [DeBERTa-Large](https://huggingface.co/microsoft/deberta-large)<sup>1</sup> | 95.5/90.1 | 90.7/88.0 | 91.3/91.1| 96.5|95.3| 69.5| 91.0| 92.6/94.6| 92.3/- |92.8/92.5 |
26
+ | [DeBERTa-XLarge](https://huggingface.co/microsoft/deberta-xlarge)<sup>1</sup> | -/- | -/- | 91.5/91.2| 97.0 | - | - | 93.1 | 92.1/94.3 | - |92.9/92.7|
27
+ | [DeBERTa-V2-XLarge](https://huggingface.co/microsoft/deberta-v2-xlarge)<sup>1</sup>|95.8/90.8| 91.4/88.9|91.7/91.6| **97.5**| 95.8|71.1|**93.9**|92.0/94.2|92.3/89.8|92.9/92.9|
28
+ |**[DeBERTa-V2-XXLarge](https://huggingface.co/microsoft/deberta-v2-xxlarge)<sup>1,2</sup>**|**96.1/91.4**|**92.2/89.7**|**91.7/91.9**|97.2|**96.0**|**72.0**| 93.5| **93.1/94.9**|**92.7/90.3** |**93.2/93.1** |
29
+ --------
30
+ #### Notes.
31
+ - <sup>1</sup> Following RoBERTa, for RTE, MRPC, STS-B, we fine-tune the tasks based on [DeBERTa-Large-MNLI](https://huggingface.co/microsoft/deberta-large-mnli), [DeBERTa-XLarge-MNLI](https://huggingface.co/microsoft/deberta-xlarge-mnli), [DeBERTa-V2-XLarge-MNLI](https://huggingface.co/microsoft/deberta-v2-xlarge-mnli), [DeBERTa-V2-XXLarge-MNLI](https://huggingface.co/microsoft/deberta-v2-xxlarge-mnli). The results of SST-2/QQP/QNLI/SQuADv2 will also be slightly improved when start from MNLI fine-tuned models, however, we only report the numbers fine-tuned from pretrained base models for those 4 tasks.
32
+ - <sup>2</sup> To try the **XXLarge** model with **[HF transformers](https://huggingface.co/transformers/main_classes/trainer.html)**, you need to specify **--sharded_ddp**
33
+
34
+ ```bash
35
+ cd transformers/examples/text-classification/
36
+ export TASK_NAME=mrpc
37
+ python -m torch.distributed.launch --nproc_per_node=8 run_glue.py --model_name_or_path microsoft/deberta-v2-xxlarge \
38
+ --task_name $TASK_NAME --do_train --do_eval --max_seq_length 128 --per_device_train_batch_size 4 \
39
+ --learning_rate 3e-6 --num_train_epochs 3 --output_dir /tmp/$TASK_NAME/ --overwrite_output_dir --sharded_ddp --fp16
40
+ ```
41
+
42
+ ### Citation
43
+
44
+ If you find DeBERTa useful for your work, please cite the following paper:
45
+
46
+ ``` latex
47
+ @inproceedings{
48
+ he2021deberta,
49
+ title={{\{}DEBERTA{\}}: {\{}DECODING{\}}-{\{}ENHANCED{\}} {\{}BERT{\}} {\{}WITH{\}} {\{}DISENTANGLED{\}} {\{}ATTENTION{\}}},
50
+ author={Pengcheng He and Xiaodong Liu and Jianfeng Gao and Weizhu Chen},
51
+ booktitle={International Conference on Learning Representations},
52
+ year={2021},
53
+ url={https://openreview.net/forum?id=XPZIaotutsD}
54
+ }
55
+ ```
config.json ADDED
@@ -0,0 +1,24 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "attention_probs_dropout_prob": 0.1,
3
+ "hidden_act": "gelu",
4
+ "hidden_dropout_prob": 0.1,
5
+ "hidden_size": 1536,
6
+ "initializer_range": 0.02,
7
+ "intermediate_size": 6144,
8
+ "max_position_embeddings": 512,
9
+ "relative_attention": true,
10
+ "position_buckets": 256,
11
+ "norm_rel_ebd": "layer_norm",
12
+ "share_att_key": true,
13
+ "pos_att_type": "p2c|c2p",
14
+ "layer_norm_eps": 1e-7,
15
+ "conv_kernel_size": 3,
16
+ "conv_act": "gelu",
17
+ "max_relative_positions": -1,
18
+ "position_biased_input": false,
19
+ "num_attention_heads": 24,
20
+ "attention_head_size": 64,
21
+ "num_hidden_layers": 48,
22
+ "type_vocab_size": 0,
23
+ "vocab_size": 128100
24
+ }
pytorch_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:421b7b47d28891477b21e8e5e97844dfcb5e03a12235dec3cad2f818dc22463b
3
+ size 3135788407
spm.model ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5598d5e96f339a8d980c15f9afd405a2e5e1be7db41de3ed13b0f03fac1e8c17
3
+ size 2447305
tokenizer_config.json ADDED
@@ -0,0 +1,4 @@
 
 
 
 
 
1
+ {
2
+ "do_lower_case": false,
3
+ "vocab_type": "spm"
4
+ }