Commit
•
29f90c4
1
Parent(s):
3c64629
Update README.md
Browse files
README.md
CHANGED
@@ -1,6 +1,7 @@
|
|
1 |
---
|
2 |
language:
|
3 |
- en
|
|
|
4 |
tags:
|
5 |
- text-classification
|
6 |
- zero-shot-classification
|
@@ -18,11 +19,14 @@ It is the only model in the model hub trained on 8 NLI datasets, including DocNL
|
|
18 |
|
19 |
The base model is [DeBERTa-v3-base from Microsoft](https://huggingface.co/microsoft/deberta-v3-base). The v3 variant of DeBERTa substantially outperforms previous versions of the model by including a different pre-training objective, see annex 11 of the original [DeBERTa paper](https://arxiv.org/pdf/2006.03654.pdf) as well as the [DeBERTa-V3 paper](https://arxiv.org/abs/2111.09543).
|
20 |
|
|
|
|
|
21 |
## Intended uses & limitations
|
22 |
#### How to use the model
|
23 |
```python
|
24 |
from transformers import AutoTokenizer, AutoModelForSequenceClassification
|
25 |
import torch
|
|
|
26 |
|
27 |
model_name = "MoritzLaurer/DeBERTa-v3-base-mnli-fever-docnli-ling-2c"
|
28 |
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
@@ -65,11 +69,11 @@ mnli-m-2c | mnli-mm-2c | fever-nli-2c | anli-all-2c | anli-r3-2c | lingnli-2c
|
|
65 |
## Limitations and bias
|
66 |
Please consult the original DeBERTa paper and literature on different NLI datasets for potential biases.
|
67 |
|
68 |
-
|
69 |
-
If you
|
70 |
|
71 |
### Ideas for cooperation or questions?
|
72 |
If you have questions or ideas for cooperation, contact me at m{dot}laurer{at}vu{dot}nl or [LinkedIn](https://www.linkedin.com/in/moritz-laurer/)
|
73 |
|
74 |
### Debugging and issues
|
75 |
-
Note that DeBERTa-v3 was released
|
|
|
1 |
---
|
2 |
language:
|
3 |
- en
|
4 |
+
license: mit
|
5 |
tags:
|
6 |
- text-classification
|
7 |
- zero-shot-classification
|
|
|
19 |
|
20 |
The base model is [DeBERTa-v3-base from Microsoft](https://huggingface.co/microsoft/deberta-v3-base). The v3 variant of DeBERTa substantially outperforms previous versions of the model by including a different pre-training objective, see annex 11 of the original [DeBERTa paper](https://arxiv.org/pdf/2006.03654.pdf) as well as the [DeBERTa-V3 paper](https://arxiv.org/abs/2111.09543).
|
21 |
|
22 |
+
For highest performance (but less speed), I recommend using https://huggingface.co/MoritzLaurer/DeBERTa-v3-large-mnli-fever-anli-ling-wanli.
|
23 |
+
|
24 |
## Intended uses & limitations
|
25 |
#### How to use the model
|
26 |
```python
|
27 |
from transformers import AutoTokenizer, AutoModelForSequenceClassification
|
28 |
import torch
|
29 |
+
device = torch.device("cuda") if torch.cuda.is_available() else torch.device("cpu")
|
30 |
|
31 |
model_name = "MoritzLaurer/DeBERTa-v3-base-mnli-fever-docnli-ling-2c"
|
32 |
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
|
|
69 |
## Limitations and bias
|
70 |
Please consult the original DeBERTa paper and literature on different NLI datasets for potential biases.
|
71 |
|
72 |
+
## Citation
|
73 |
+
If you use this model, please cite: Laurer, Moritz, Wouter van Atteveldt, Andreu Salleras Casas, and Kasper Welbers. 2022. ‘Less Annotating, More Classifying – Addressing the Data Scarcity Issue of Supervised Machine Learning with Deep Transfer Learning and BERT - NLI’. Preprint, June. Open Science Framework. https://osf.io/74b8k.
|
74 |
|
75 |
### Ideas for cooperation or questions?
|
76 |
If you have questions or ideas for cooperation, contact me at m{dot}laurer{at}vu{dot}nl or [LinkedIn](https://www.linkedin.com/in/moritz-laurer/)
|
77 |
|
78 |
### Debugging and issues
|
79 |
+
Note that DeBERTa-v3 was released on 06.12.21 and older versions of HF Transformers seem to have issues running the model (e.g. resulting in an issue with the tokenizer). Using Transformers>=4.13 might solve some issues.
|