File size: 3,484 Bytes
586a45b
ff8a366
 
 
fde7023
 
ff8a366
 
fde7023
 
 
586a45b
fde7023
 
 
 
 
 
 
 
 
 
 
 
293b6a9
 
fde7023
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
293b6a9
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
e12b9e3
d92ed9c
e12b9e3
4ee38d0
e12b9e3
a7e7c80
e12b9e3
 
a7e7c80
 
e12b9e3
fde7023
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
---
language:
- en
license: mit
datasets:
- glue
pipeline_tag: zero-shot-classification
base_model: intfloat/e5-large
model-index:
- name: e5-large-mnli
  results: []
---

# e5-large-mnli

This model is a fine-tuned version of [intfloat/e5-large](https://huggingface.co/intfloat/e5-large) on the glue dataset.

## Model description

[Text Embeddings by Weakly-Supervised Contrastive Pre-training](https://arxiv.org/pdf/2212.03533.pdf).
Liang Wang, Nan Yang, Xiaolong Huang, Binxing Jiao, Linjun Yang, Daxin Jiang, Rangan Majumder, Furu Wei, arXiv 2022

## How to use the model

### With the zero-shot classification pipeline

The model can be loaded with the `zero-shot-classification` pipeline like so:

```python
from transformers import pipeline
classifier = pipeline("zero-shot-classification",
                      model="mjwong/e5-large-mnli")
```

You can then use this pipeline to classify sequences into any of the class names you specify.

```python
sequence_to_classify = "one day I will see the world"
candidate_labels = ['travel', 'cooking', 'dancing']
classifier(sequence_to_classify, candidate_labels)
```

If more than one candidate label can be correct, pass `multi_class=True` to calculate each class independently:

```python
candidate_labels = ['travel', 'cooking', 'dancing', 'exploration']
classifier(sequence_to_classify, candidate_labels, multi_class=True)
```

### With manual PyTorch

The model can also be applied on NLI tasks like so:

```python
import torch
from transformers import AutoTokenizer, AutoModelForSequenceClassification

# device = "cuda:0" or "cpu"
device = torch.device("cuda") if torch.cuda.is_available() else torch.device("cpu")

model_name = "mjwong/e5-large-mnli"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForSequenceClassification.from_pretrained(model_name)

premise = "But I thought you'd sworn off coffee."
hypothesis = "I thought that you vowed to drink more coffee."

input = tokenizer(premise, hypothesis, truncation=True, return_tensors="pt")
output = model(input["input_ids"].to(device))
prediction = torch.softmax(output["logits"][0], -1).tolist()
label_names = ["entailment", "neutral", "contradiction"]
prediction = {name: round(float(pred) * 100, 2) for pred, name in zip(prediction, label_names)}
print(prediction)
```

### Eval results
The model was evaluated using the dev sets for MultiNLI and test sets for ANLI. The metric used is accuracy.

|Datasets|mnli_dev_m|mnli_dev_mm|anli_test_r1|anli_test_r2|anli_test_r3|
| :---: | :---: | :---: | :---: | :---: | :---: |
|[e5-base-v2-mnli-anli](https://huggingface.co/mjwong/e5-base-v2-mnli-anli)|0.812|0.809|0.557|0.460|0.448|
|[e5-large-mnli](https://huggingface.co/mjwong/e5-large-mnli)|0.868|0.869|0.301|0.296|0.294|
|[e5-large-mnli-anli](https://huggingface.co/mjwong/e5-large-mnli-anli)|0.843|0.848|0.646|0.484|0.458|
|[e5-large-v2-mnli](https://huggingface.co/mjwong/e5-large-v2-mnli)|0.875|0.876|0.354|0.298|0.313|
|[e5-large-v2-mnli-anli](https://huggingface.co/mjwong/e5-large-v2-mnli-anli)|0.846|0.848|0.638|0.474|0.479|

### Training hyperparameters

The following hyperparameters were used during training:

- learning_rate: 2e-05
- train_batch_size: 16
- eval_batch_size: 16
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_ratio: 0.1
- num_epochs: 2

### Framework versions
- Transformers 4.28.1
- Pytorch 1.12.1+cu116
- Datasets 2.11.0
- Tokenizers 0.12.1