Initial commit
Browse files- 1_Pooling/config.json +7 -0
- README.md +136 -0
- config.json +23 -0
- config_sentence_transformers.json +7 -0
- model.safetensors +3 -0
- modules.json +14 -0
- sentence_bert_config.json +4 -0
- special_tokens_map.json +37 -0
- tokenizer_config.json +63 -0
- vocab.txt +0 -0
1_Pooling/config.json
ADDED
@@ -0,0 +1,7 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"word_embedding_dimension": 768,
|
3 |
+
"pooling_mode_cls_token": true,
|
4 |
+
"pooling_mode_mean_tokens": false,
|
5 |
+
"pooling_mode_max_tokens": false,
|
6 |
+
"pooling_mode_mean_sqrt_len_tokens": false
|
7 |
+
}
|
README.md
ADDED
@@ -0,0 +1,136 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
language:
|
3 |
+
- ja
|
4 |
+
pipeline_tag: sentence-similarity
|
5 |
+
tags:
|
6 |
+
- sentence-transformers
|
7 |
+
- feature-extraction
|
8 |
+
- sentence-similarity
|
9 |
+
- transformers
|
10 |
+
|
11 |
+
---
|
12 |
+
|
13 |
+
# fio-base-japanese-v0.1
|
14 |
+
|
15 |
+
日本語版は近日公開予定です(日本語を勉強中なので、間違いはご容赦ください!)
|
16 |
+
|
17 |
+
fio-base-japanese-v0.1 is a proof of concept, and the first release of the Fio family of Japanese embeddings. It is based on [cl-tohoku/bert-base-japanese-v3](https://huggingface.co/cl-tohoku/bert-base-japanese-v3) and trained on limited volumes of data on single GPU.
|
18 |
+
|
19 |
+
For more information, please refer to [my notes on Fio](https://ben.clavie.eu/fio).
|
20 |
+
|
21 |
+
#### Datasets
|
22 |
+
|
23 |
+
Similarity/Entailment:
|
24 |
+
- JSTS (train)
|
25 |
+
- JSNLI (train)
|
26 |
+
- JNLI (train)
|
27 |
+
- JSICK (train)
|
28 |
+
|
29 |
+
Retrieval:
|
30 |
+
- MMARCO (Multilingual Marco) (train, 124k sentence pairs, ~<2% of the full data)
|
31 |
+
- Mr.TyDI (train)
|
32 |
+
- MIRACL (train, 50% sample)
|
33 |
+
- ~~JSQuAD (train, 50% sample, no LLM enhancement)~~ JSQuAD is not used in the released version, to serve as an unseen test set.
|
34 |
+
|
35 |
+
#### Results
|
36 |
+
|
37 |
+
This is adapted and truncated (to keep only the most popular models) from [oshizo's benchmarking github repo](https://github.com/oshizo/JapaneseEmbeddingEval), please check it out for more information and give it a star as it was very useful!
|
38 |
+
|
39 |
+
Italic denotes best model for its size (base/large | 768/1024), bold denotes best overall.
|
40 |
+
|
41 |
+
| Model | JSTS valid-v1.1 | JSICK test | MIRACL dev | Average |
|
42 |
+
|-------------------------------------------------|-----------------|------------|------------|---------|
|
43 |
+
| bclavie/fio-base-japanese-v0.1 | **_0.863_** | **_0.894_** | 0.715 | _0.824_ |
|
44 |
+
| cl-nagoya/sup-simcse-ja-base | 0.809 | 0.827 | 0.527 | 0.721 |
|
45 |
+
| cl-nagoya/sup-simcse-ja-large | 0.831 | 0.831 | 0.507 | 0.723 |
|
46 |
+
| colorfulscoop/sbert-base-ja | 0.742 | 0.657 | 0.254 | 0.551 |
|
47 |
+
| intfloat/multilingual-e5-base | 0.796 | 0.806 | **0.845** | 0.816 |
|
48 |
+
| intfloat/multilingual-e5-large | 0.819 | 0.794 | **0.883** | **_0.832_** |
|
49 |
+
| pkshatech/GLuCoSE-base-ja | 0.818 | 0.757 | 0.692 | 0.755 |
|
50 |
+
| text-embedding-ada-002 | 0.790 | 0.789 | 0.7232 | 0.768 |
|
51 |
+
|
52 |
+
|
53 |
+
## Usage (Sentence-Transformers)
|
54 |
+
|
55 |
+
This model is best used through [sentence-transformers](https://www.SBERT.net). If you don't have it, it's easy to install:
|
56 |
+
|
57 |
+
```
|
58 |
+
pip install -U sentence-transformers
|
59 |
+
```
|
60 |
+
|
61 |
+
Then you can use the model like this:
|
62 |
+
|
63 |
+
```python
|
64 |
+
from sentence_transformers import SentenceTransformer
|
65 |
+
sentences = ["こんにちは、世界!", "文埋め込み最高!文埋め込み最高と叫びなさい", "極度乾燥しなさい"]
|
66 |
+
|
67 |
+
model = SentenceTransformer('bclavie/fio-base-japanese-v0.1')
|
68 |
+
embeddings = model.encode(sentences)
|
69 |
+
print(embeddings)
|
70 |
+
```
|
71 |
+
|
72 |
+
|
73 |
+
## Usage
|
74 |
+
|
75 |
+
If using for a retrieval task, you must prefix your query with `"関連記事を取得するために使用できるこの文の表現を生成します: "`.
|
76 |
+
|
77 |
+
### Usage (HuggingFace Transformers)
|
78 |
+
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.
|
79 |
+
|
80 |
+
```python
|
81 |
+
from transformers import AutoTokenizer, AutoModel
|
82 |
+
import torch
|
83 |
+
|
84 |
+
|
85 |
+
def cls_pooling(model_output, attention_mask):
|
86 |
+
return model_output[0][:,0]
|
87 |
+
|
88 |
+
|
89 |
+
# Sentences we want sentence embeddings for
|
90 |
+
sentences = ['This is an example sentence', 'Each sentence is converted']
|
91 |
+
|
92 |
+
# Load model from HuggingFace Hub
|
93 |
+
tokenizer = AutoTokenizer.from_pretrained('{MODEL_NAME}')
|
94 |
+
model = AutoModel.from_pretrained('{MODEL_NAME}')
|
95 |
+
|
96 |
+
# Tokenize sentences
|
97 |
+
encoded_input = tokenizer(sentences, padding=True, truncation=True, return_tensors='pt')
|
98 |
+
|
99 |
+
# Compute token embeddings
|
100 |
+
with torch.no_grad():
|
101 |
+
model_output = model(**encoded_input)
|
102 |
+
|
103 |
+
# Perform pooling. In this case, cls pooling.
|
104 |
+
sentence_embeddings = cls_pooling(model_output, encoded_input['attention_mask'])
|
105 |
+
|
106 |
+
print("Sentence embeddings:")
|
107 |
+
print(sentence_embeddings)
|
108 |
+
```
|
109 |
+
|
110 |
+
|
111 |
+
|
112 |
+
### Evaluation Results
|
113 |
+
|
114 |
+
<!--- Describe how your model was evaluated -->
|
115 |
+
|
116 |
+
For an automated evaluation of this model, see the *Sentence Embeddings Benchmark*: [https://seb.sbert.net](https://seb.sbert.net?model_name={MODEL_NAME})
|
117 |
+
|
118 |
+
|
119 |
+
|
120 |
+
## Full Model Architecture
|
121 |
+
```
|
122 |
+
SentenceTransformer(
|
123 |
+
(0): Transformer({'max_seq_length': 512, 'do_lower_case': False}) with Transformer model: BertModel
|
124 |
+
(1): Pooling({'word_embedding_dimension': 768, 'pooling_mode_cls_token': True, 'pooling_mode_mean_tokens': False, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False})
|
125 |
+
)
|
126 |
+
```
|
127 |
+
|
128 |
+
## Citing & Authors
|
129 |
+
|
130 |
+
@misc{
|
131 |
+
bclavie-fio-embeddings,
|
132 |
+
author = {Benjamin Clavié},
|
133 |
+
title = {Fio Japanese Embeddings},
|
134 |
+
year = {2023},
|
135 |
+
howpublished = {\url{https://ben.clavie.eu/fio}}
|
136 |
+
}
|
config.json
ADDED
@@ -0,0 +1,23 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"_name_or_path": "fio-base-japanese-v0.1",
|
3 |
+
"architectures": ["BertModel"],
|
4 |
+
"attention_probs_dropout_prob": 0.1,
|
5 |
+
"classifier_dropout": null,
|
6 |
+
"hidden_act": "gelu",
|
7 |
+
"hidden_dropout_prob": 0.1,
|
8 |
+
"hidden_size": 768,
|
9 |
+
"initializer_range": 0.02,
|
10 |
+
"intermediate_size": 3072,
|
11 |
+
"layer_norm_eps": 1e-12,
|
12 |
+
"max_position_embeddings": 512,
|
13 |
+
"model_type": "bert",
|
14 |
+
"num_attention_heads": 12,
|
15 |
+
"num_hidden_layers": 12,
|
16 |
+
"pad_token_id": 0,
|
17 |
+
"position_embedding_type": "absolute",
|
18 |
+
"torch_dtype": "float32",
|
19 |
+
"transformers_version": "4.36.1",
|
20 |
+
"type_vocab_size": 2,
|
21 |
+
"use_cache": false,
|
22 |
+
"vocab_size": 32768
|
23 |
+
}
|
config_sentence_transformers.json
ADDED
@@ -0,0 +1,7 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"__version__": {
|
3 |
+
"sentence_transformers": "2.2.2",
|
4 |
+
"transformers": "4.36.1",
|
5 |
+
"pytorch": "2.1.0+cu118"
|
6 |
+
}
|
7 |
+
}
|
model.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:d19517586fff72065b7577fc715463f433876b7b40b96fd5e4eb3a16f626f663
|
3 |
+
size 444851048
|
modules.json
ADDED
@@ -0,0 +1,14 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
[
|
2 |
+
{
|
3 |
+
"idx": 0,
|
4 |
+
"name": "0",
|
5 |
+
"path": "",
|
6 |
+
"type": "sentence_transformers.models.Transformer"
|
7 |
+
},
|
8 |
+
{
|
9 |
+
"idx": 1,
|
10 |
+
"name": "1",
|
11 |
+
"path": "1_Pooling",
|
12 |
+
"type": "sentence_transformers.models.Pooling"
|
13 |
+
}
|
14 |
+
]
|
sentence_bert_config.json
ADDED
@@ -0,0 +1,4 @@
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"max_seq_length": 512,
|
3 |
+
"do_lower_case": false
|
4 |
+
}
|
special_tokens_map.json
ADDED
@@ -0,0 +1,37 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"cls_token": {
|
3 |
+
"content": "[CLS]",
|
4 |
+
"lstrip": false,
|
5 |
+
"normalized": false,
|
6 |
+
"rstrip": false,
|
7 |
+
"single_word": false
|
8 |
+
},
|
9 |
+
"mask_token": {
|
10 |
+
"content": "[MASK]",
|
11 |
+
"lstrip": false,
|
12 |
+
"normalized": false,
|
13 |
+
"rstrip": false,
|
14 |
+
"single_word": false
|
15 |
+
},
|
16 |
+
"pad_token": {
|
17 |
+
"content": "[PAD]",
|
18 |
+
"lstrip": false,
|
19 |
+
"normalized": false,
|
20 |
+
"rstrip": false,
|
21 |
+
"single_word": false
|
22 |
+
},
|
23 |
+
"sep_token": {
|
24 |
+
"content": "[SEP]",
|
25 |
+
"lstrip": false,
|
26 |
+
"normalized": false,
|
27 |
+
"rstrip": false,
|
28 |
+
"single_word": false
|
29 |
+
},
|
30 |
+
"unk_token": {
|
31 |
+
"content": "[UNK]",
|
32 |
+
"lstrip": false,
|
33 |
+
"normalized": false,
|
34 |
+
"rstrip": false,
|
35 |
+
"single_word": false
|
36 |
+
}
|
37 |
+
}
|
tokenizer_config.json
ADDED
@@ -0,0 +1,63 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"added_tokens_decoder": {
|
3 |
+
"0": {
|
4 |
+
"content": "[PAD]",
|
5 |
+
"lstrip": false,
|
6 |
+
"normalized": false,
|
7 |
+
"rstrip": false,
|
8 |
+
"single_word": false,
|
9 |
+
"special": true
|
10 |
+
},
|
11 |
+
"1": {
|
12 |
+
"content": "[UNK]",
|
13 |
+
"lstrip": false,
|
14 |
+
"normalized": false,
|
15 |
+
"rstrip": false,
|
16 |
+
"single_word": false,
|
17 |
+
"special": true
|
18 |
+
},
|
19 |
+
"2": {
|
20 |
+
"content": "[CLS]",
|
21 |
+
"lstrip": false,
|
22 |
+
"normalized": false,
|
23 |
+
"rstrip": false,
|
24 |
+
"single_word": false,
|
25 |
+
"special": true
|
26 |
+
},
|
27 |
+
"3": {
|
28 |
+
"content": "[SEP]",
|
29 |
+
"lstrip": false,
|
30 |
+
"normalized": false,
|
31 |
+
"rstrip": false,
|
32 |
+
"single_word": false,
|
33 |
+
"special": true
|
34 |
+
},
|
35 |
+
"4": {
|
36 |
+
"content": "[MASK]",
|
37 |
+
"lstrip": false,
|
38 |
+
"normalized": false,
|
39 |
+
"rstrip": false,
|
40 |
+
"single_word": false,
|
41 |
+
"special": true
|
42 |
+
}
|
43 |
+
},
|
44 |
+
"clean_up_tokenization_spaces": true,
|
45 |
+
"cls_token": "[CLS]",
|
46 |
+
"do_lower_case": false,
|
47 |
+
"do_subword_tokenize": true,
|
48 |
+
"do_word_tokenize": true,
|
49 |
+
"jumanpp_kwargs": null,
|
50 |
+
"mask_token": "[MASK]",
|
51 |
+
"mecab_kwargs": {
|
52 |
+
"mecab_dic": "unidic_lite"
|
53 |
+
},
|
54 |
+
"model_max_length": 512,
|
55 |
+
"never_split": null,
|
56 |
+
"pad_token": "[PAD]",
|
57 |
+
"sep_token": "[SEP]",
|
58 |
+
"subword_tokenizer_type": "wordpiece",
|
59 |
+
"sudachi_kwargs": null,
|
60 |
+
"tokenizer_class": "BertJapaneseTokenizer",
|
61 |
+
"unk_token": "[UNK]",
|
62 |
+
"word_tokenizer_type": "mecab"
|
63 |
+
}
|
vocab.txt
ADDED
The diff for this file is too large to render.
See raw diff
|
|