luismsgomes
commited on
added bibtex
Browse files
README.md
CHANGED
@@ -1,132 +1,145 @@
|
|
1 |
-
---
|
2 |
-
language: pt
|
3 |
-
license: mit
|
4 |
-
library_name: sentence-transformers
|
5 |
-
pipeline_tag: sentence-similarity
|
6 |
-
tags:
|
7 |
-
- sentence-transformers
|
8 |
-
- feature-extraction
|
9 |
-
- sentence-similarity
|
10 |
-
- transformers
|
11 |
-
|
12 |
-
---
|
13 |
-
|
14 |
-
# Serafim 900m Portuguese (PT) Sentence Encoder
|
15 |
-
|
16 |
-
This is a [sentence-transformers](https://www.SBERT.net) model: It maps sentences & paragraphs to a 1536 dimensional dense vector space and can be used for tasks like clustering or semantic search.
|
17 |
-
|
18 |
-
<!--- Describe your model here -->
|
19 |
-
|
20 |
-
## Usage (Sentence-Transformers)
|
21 |
-
|
22 |
-
Using this model becomes easy when you have [sentence-transformers](https://www.SBERT.net) installed:
|
23 |
-
|
24 |
-
```
|
25 |
-
pip install -U sentence-transformers
|
26 |
-
```
|
27 |
-
|
28 |
-
Then you can use the model like this:
|
29 |
-
|
30 |
-
```python
|
31 |
-
from sentence_transformers import SentenceTransformer
|
32 |
-
sentences = ["This is an example sentence", "Each sentence is converted"]
|
33 |
-
|
34 |
-
model = SentenceTransformer('PORTULAN/serafim-900m-portuguese-pt-sentence-encoder')
|
35 |
-
embeddings = model.encode(sentences)
|
36 |
-
print(embeddings)
|
37 |
-
```
|
38 |
-
|
39 |
-
|
40 |
-
|
41 |
-
## Usage (HuggingFace Transformers)
|
42 |
-
Without [sentence-transformers](https://www.SBERT.net), you can use the model like this: First, you pass your input through the transformer model, then you have to apply the right pooling-operation on-top of the contextualized word embeddings.
|
43 |
-
|
44 |
-
```python
|
45 |
-
from transformers import AutoTokenizer, AutoModel
|
46 |
-
import torch
|
47 |
-
|
48 |
-
|
49 |
-
#Mean Pooling - Take attention mask into account for correct averaging
|
50 |
-
def mean_pooling(model_output, attention_mask):
|
51 |
-
token_embeddings = model_output[0] #First element of model_output contains all token embeddings
|
52 |
-
input_mask_expanded = attention_mask.unsqueeze(-1).expand(token_embeddings.size()).float()
|
53 |
-
return torch.sum(token_embeddings * input_mask_expanded, 1) / torch.clamp(input_mask_expanded.sum(1), min=1e-9)
|
54 |
-
|
55 |
-
|
56 |
-
# Sentences we want sentence embeddings for
|
57 |
-
sentences = ['This is an example sentence', 'Each sentence is converted']
|
58 |
-
|
59 |
-
# Load model from HuggingFace Hub
|
60 |
-
tokenizer = AutoTokenizer.from_pretrained('PORTULAN/serafim-900m-portuguese-pt-sentence-encoder')
|
61 |
-
model = AutoModel.from_pretrained('PORTULAN/serafim-900m-portuguese-pt-sentence-encoder')
|
62 |
-
|
63 |
-
# Tokenize sentences
|
64 |
-
encoded_input = tokenizer(sentences, padding=True, truncation=True, return_tensors='pt')
|
65 |
-
|
66 |
-
# Compute token embeddings
|
67 |
-
with torch.no_grad():
|
68 |
-
model_output = model(**encoded_input)
|
69 |
-
|
70 |
-
# Perform pooling. In this case, mean pooling.
|
71 |
-
sentence_embeddings = mean_pooling(model_output, encoded_input['attention_mask'])
|
72 |
-
|
73 |
-
print("Sentence embeddings:")
|
74 |
-
print(sentence_embeddings)
|
75 |
-
```
|
76 |
-
|
77 |
-
|
78 |
-
|
79 |
-
## Evaluation Results
|
80 |
-
|
81 |
-
<!--- Describe how your model was evaluated -->
|
82 |
-
|
83 |
-
For an automated evaluation of this model, see the *Sentence Embeddings Benchmark*: [https://seb.sbert.net](https://seb.sbert.net?model_name=PORTULAN/serafim-900m-portuguese-pt-sentence-encoder)
|
84 |
-
|
85 |
-
|
86 |
-
## Training
|
87 |
-
The model was trained with the parameters:
|
88 |
-
|
89 |
-
**DataLoader**:
|
90 |
-
|
91 |
-
`torch.utils.data.dataloader.DataLoader` of length 1183 with parameters:
|
92 |
-
```
|
93 |
-
{'batch_size': 16, 'sampler': 'torch.utils.data.sampler.RandomSampler', 'batch_sampler': 'torch.utils.data.sampler.BatchSampler'}
|
94 |
-
```
|
95 |
-
|
96 |
-
**Loss**:
|
97 |
-
|
98 |
-
`sentence_transformers.losses.CoSENTLoss.CoSENTLoss` with parameters:
|
99 |
-
```
|
100 |
-
{'scale': 20.0, 'similarity_fct': 'pairwise_cos_sim'}
|
101 |
-
```
|
102 |
-
|
103 |
-
Parameters of the fit()-Method:
|
104 |
-
```
|
105 |
-
{
|
106 |
-
"epochs": 10,
|
107 |
-
"evaluation_steps": 119,
|
108 |
-
"evaluator": "sentence_transformers.evaluation.EmbeddingSimilarityEvaluator.EmbeddingSimilarityEvaluator",
|
109 |
-
"max_grad_norm": 1,
|
110 |
-
"optimizer_class": "<class 'torch.optim.adamw.AdamW'>",
|
111 |
-
"optimizer_params": {
|
112 |
-
"lr": 1e-06
|
113 |
-
},
|
114 |
-
"scheduler": "WarmupLinear",
|
115 |
-
"steps_per_epoch": 1183,
|
116 |
-
"warmup_steps": 1183,
|
117 |
-
"weight_decay": 0.01
|
118 |
-
}
|
119 |
-
```
|
120 |
-
|
121 |
-
|
122 |
-
## Full Model Architecture
|
123 |
-
```
|
124 |
-
SentenceTransformer(
|
125 |
-
(0): Transformer({'max_seq_length': 128, 'do_lower_case': False}) with Transformer model: DebertaV2Model
|
126 |
-
(1): Pooling({'word_embedding_dimension': 1536, 'pooling_mode_cls_token': False, 'pooling_mode_mean_tokens': True, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False, 'pooling_mode_weightedmean_tokens': False, 'pooling_mode_lasttoken': False, 'include_prompt': True})
|
127 |
-
)
|
128 |
-
```
|
129 |
-
|
130 |
-
## Citing & Authors
|
131 |
-
|
132 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
language: pt
|
3 |
+
license: mit
|
4 |
+
library_name: sentence-transformers
|
5 |
+
pipeline_tag: sentence-similarity
|
6 |
+
tags:
|
7 |
+
- sentence-transformers
|
8 |
+
- feature-extraction
|
9 |
+
- sentence-similarity
|
10 |
+
- transformers
|
11 |
+
|
12 |
+
---
|
13 |
+
|
14 |
+
# Serafim 900m Portuguese (PT) Sentence Encoder
|
15 |
+
|
16 |
+
This is a [sentence-transformers](https://www.SBERT.net) model: It maps sentences & paragraphs to a 1536 dimensional dense vector space and can be used for tasks like clustering or semantic search.
|
17 |
+
|
18 |
+
<!--- Describe your model here -->
|
19 |
+
|
20 |
+
## Usage (Sentence-Transformers)
|
21 |
+
|
22 |
+
Using this model becomes easy when you have [sentence-transformers](https://www.SBERT.net) installed:
|
23 |
+
|
24 |
+
```
|
25 |
+
pip install -U sentence-transformers
|
26 |
+
```
|
27 |
+
|
28 |
+
Then you can use the model like this:
|
29 |
+
|
30 |
+
```python
|
31 |
+
from sentence_transformers import SentenceTransformer
|
32 |
+
sentences = ["This is an example sentence", "Each sentence is converted"]
|
33 |
+
|
34 |
+
model = SentenceTransformer('PORTULAN/serafim-900m-portuguese-pt-sentence-encoder')
|
35 |
+
embeddings = model.encode(sentences)
|
36 |
+
print(embeddings)
|
37 |
+
```
|
38 |
+
|
39 |
+
|
40 |
+
|
41 |
+
## Usage (HuggingFace Transformers)
|
42 |
+
Without [sentence-transformers](https://www.SBERT.net), you can use the model like this: First, you pass your input through the transformer model, then you have to apply the right pooling-operation on-top of the contextualized word embeddings.
|
43 |
+
|
44 |
+
```python
|
45 |
+
from transformers import AutoTokenizer, AutoModel
|
46 |
+
import torch
|
47 |
+
|
48 |
+
|
49 |
+
#Mean Pooling - Take attention mask into account for correct averaging
|
50 |
+
def mean_pooling(model_output, attention_mask):
|
51 |
+
token_embeddings = model_output[0] #First element of model_output contains all token embeddings
|
52 |
+
input_mask_expanded = attention_mask.unsqueeze(-1).expand(token_embeddings.size()).float()
|
53 |
+
return torch.sum(token_embeddings * input_mask_expanded, 1) / torch.clamp(input_mask_expanded.sum(1), min=1e-9)
|
54 |
+
|
55 |
+
|
56 |
+
# Sentences we want sentence embeddings for
|
57 |
+
sentences = ['This is an example sentence', 'Each sentence is converted']
|
58 |
+
|
59 |
+
# Load model from HuggingFace Hub
|
60 |
+
tokenizer = AutoTokenizer.from_pretrained('PORTULAN/serafim-900m-portuguese-pt-sentence-encoder')
|
61 |
+
model = AutoModel.from_pretrained('PORTULAN/serafim-900m-portuguese-pt-sentence-encoder')
|
62 |
+
|
63 |
+
# Tokenize sentences
|
64 |
+
encoded_input = tokenizer(sentences, padding=True, truncation=True, return_tensors='pt')
|
65 |
+
|
66 |
+
# Compute token embeddings
|
67 |
+
with torch.no_grad():
|
68 |
+
model_output = model(**encoded_input)
|
69 |
+
|
70 |
+
# Perform pooling. In this case, mean pooling.
|
71 |
+
sentence_embeddings = mean_pooling(model_output, encoded_input['attention_mask'])
|
72 |
+
|
73 |
+
print("Sentence embeddings:")
|
74 |
+
print(sentence_embeddings)
|
75 |
+
```
|
76 |
+
|
77 |
+
|
78 |
+
|
79 |
+
## Evaluation Results
|
80 |
+
|
81 |
+
<!--- Describe how your model was evaluated -->
|
82 |
+
|
83 |
+
For an automated evaluation of this model, see the *Sentence Embeddings Benchmark*: [https://seb.sbert.net](https://seb.sbert.net?model_name=PORTULAN/serafim-900m-portuguese-pt-sentence-encoder)
|
84 |
+
|
85 |
+
|
86 |
+
## Training
|
87 |
+
The model was trained with the parameters:
|
88 |
+
|
89 |
+
**DataLoader**:
|
90 |
+
|
91 |
+
`torch.utils.data.dataloader.DataLoader` of length 1183 with parameters:
|
92 |
+
```
|
93 |
+
{'batch_size': 16, 'sampler': 'torch.utils.data.sampler.RandomSampler', 'batch_sampler': 'torch.utils.data.sampler.BatchSampler'}
|
94 |
+
```
|
95 |
+
|
96 |
+
**Loss**:
|
97 |
+
|
98 |
+
`sentence_transformers.losses.CoSENTLoss.CoSENTLoss` with parameters:
|
99 |
+
```
|
100 |
+
{'scale': 20.0, 'similarity_fct': 'pairwise_cos_sim'}
|
101 |
+
```
|
102 |
+
|
103 |
+
Parameters of the fit()-Method:
|
104 |
+
```
|
105 |
+
{
|
106 |
+
"epochs": 10,
|
107 |
+
"evaluation_steps": 119,
|
108 |
+
"evaluator": "sentence_transformers.evaluation.EmbeddingSimilarityEvaluator.EmbeddingSimilarityEvaluator",
|
109 |
+
"max_grad_norm": 1,
|
110 |
+
"optimizer_class": "<class 'torch.optim.adamw.AdamW'>",
|
111 |
+
"optimizer_params": {
|
112 |
+
"lr": 1e-06
|
113 |
+
},
|
114 |
+
"scheduler": "WarmupLinear",
|
115 |
+
"steps_per_epoch": 1183,
|
116 |
+
"warmup_steps": 1183,
|
117 |
+
"weight_decay": 0.01
|
118 |
+
}
|
119 |
+
```
|
120 |
+
|
121 |
+
|
122 |
+
## Full Model Architecture
|
123 |
+
```
|
124 |
+
SentenceTransformer(
|
125 |
+
(0): Transformer({'max_seq_length': 128, 'do_lower_case': False}) with Transformer model: DebertaV2Model
|
126 |
+
(1): Pooling({'word_embedding_dimension': 1536, 'pooling_mode_cls_token': False, 'pooling_mode_mean_tokens': True, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False, 'pooling_mode_weightedmean_tokens': False, 'pooling_mode_lasttoken': False, 'include_prompt': True})
|
127 |
+
)
|
128 |
+
```
|
129 |
+
|
130 |
+
## Citing & Authors
|
131 |
+
|
132 |
+
The article has been presented at EPIA 2024 conference but the Springer proceedings are not available yet.
|
133 |
+
In the meantime, if you use this model you may cite the arXiv preprint:
|
134 |
+
|
135 |
+
@misc{gomes2024opensentenceembeddingsportuguese,
|
136 |
+
title={Open Sentence Embeddings for Portuguese with the Serafim PT* encoders family},
|
137 |
+
author={Lu铆s Gomes and Ant贸nio Branco and Jo茫o Silva and Jo茫o Rodrigues and Rodrigo Santos},
|
138 |
+
year={2024},
|
139 |
+
eprint={2407.19527},
|
140 |
+
archivePrefix={arXiv},
|
141 |
+
primaryClass={cs.CL},
|
142 |
+
url={https://arxiv.org/abs/2407.19527},
|
143 |
+
}
|
144 |
+
|
145 |
+
|