Files changed (1) hide show
  1. README.md +90 -141
README.md CHANGED
@@ -1,199 +1,148 @@
1
  ---
2
  library_name: transformers
3
- tags: []
 
 
 
4
  ---
5
 
6
- # Model Card for Model ID
7
-
8
- <!-- Provide a quick summary of what the model is/does. -->
9
-
10
-
11
 
12
  ## Model Details
13
 
14
  ### Model Description
15
 
16
- <!-- Provide a longer summary of what this model is. -->
17
-
18
- This is the model card of a 🤗 transformers model that has been pushed on the Hub. This model card has been automatically generated.
19
 
20
- - **Developed by:** [More Information Needed]
21
- - **Funded by [optional]:** [More Information Needed]
22
- - **Shared by [optional]:** [More Information Needed]
23
- - **Model type:** [More Information Needed]
24
- - **Language(s) (NLP):** [More Information Needed]
25
- - **License:** [More Information Needed]
26
- - **Finetuned from model [optional]:** [More Information Needed]
27
 
28
- ### Model Sources [optional]
 
 
 
 
 
 
 
29
 
30
- <!-- Provide the basic links for the model. -->
 
 
 
 
31
 
32
- - **Repository:** [More Information Needed]
33
- - **Paper [optional]:** [More Information Needed]
34
- - **Demo [optional]:** [More Information Needed]
35
 
36
  ## Uses
37
 
38
- <!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
 
 
39
 
40
  ### Direct Use
41
 
42
- <!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
43
-
44
- [More Information Needed]
45
-
46
- ### Downstream Use [optional]
47
-
48
- <!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
49
-
50
- [More Information Needed]
51
-
52
- ### Out-of-Scope Use
53
-
54
- <!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
55
-
56
- [More Information Needed]
57
-
58
- ## Bias, Risks, and Limitations
59
-
60
- <!-- This section is meant to convey both technical and sociotechnical limitations. -->
61
 
62
- [More Information Needed]
 
63
 
64
- ### Recommendations
 
 
 
65
 
66
- <!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
 
 
67
 
68
- Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
69
 
70
- ## How to Get Started with the Model
 
 
71
 
72
- Use the code below to get started with the model.
73
-
74
- [More Information Needed]
75
 
76
  ## Training Details
77
 
78
  ### Training Data
 
 
 
 
 
 
79
 
80
- <!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
81
-
82
- [More Information Needed]
83
 
84
  ### Training Procedure
 
 
85
 
86
- <!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
87
-
88
- #### Preprocessing [optional]
89
-
90
- [More Information Needed]
91
-
92
 
93
  #### Training Hyperparameters
 
94
 
95
- - **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
96
-
97
- #### Speeds, Sizes, Times [optional]
98
-
99
- <!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
100
-
101
- [More Information Needed]
102
 
103
  ## Evaluation
 
 
104
 
105
- <!-- This section describes the evaluation protocols and provides the results. -->
106
-
107
- ### Testing Data, Factors & Metrics
108
-
109
- #### Testing Data
110
-
111
- <!-- This should link to a Dataset Card if possible. -->
112
-
113
- [More Information Needed]
114
-
115
- #### Factors
116
 
117
- <!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
118
 
119
- [More Information Needed]
120
 
121
- #### Metrics
 
122
 
123
- <!-- These are the evaluation metrics being used, ideally with a description of why. -->
 
 
 
 
124
 
125
- [More Information Needed]
 
 
 
126
 
127
- ### Results
128
-
129
- [More Information Needed]
130
-
131
- #### Summary
132
-
133
-
134
-
135
- ## Model Examination [optional]
136
-
137
- <!-- Relevant interpretability work for the model goes here -->
138
-
139
- [More Information Needed]
140
-
141
- ## Environmental Impact
142
-
143
- <!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
144
-
145
- Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
146
-
147
- - **Hardware Type:** [More Information Needed]
148
- - **Hours used:** [More Information Needed]
149
- - **Cloud Provider:** [More Information Needed]
150
- - **Compute Region:** [More Information Needed]
151
- - **Carbon Emitted:** [More Information Needed]
152
-
153
- ## Technical Specifications [optional]
154
-
155
- ### Model Architecture and Objective
156
-
157
- [More Information Needed]
158
 
159
  ### Compute Infrastructure
160
 
161
- [More Information Needed]
162
-
163
- #### Hardware
164
 
165
- [More Information Needed]
166
 
167
  #### Software
168
 
169
- [More Information Needed]
170
-
171
- ## Citation [optional]
172
-
173
- <!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
174
-
175
- **BibTeX:**
176
-
177
- [More Information Needed]
178
 
179
- **APA:**
180
 
181
- [More Information Needed]
182
 
183
- ## Glossary [optional]
184
 
185
- <!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
186
-
187
- [More Information Needed]
188
-
189
- ## More Information [optional]
190
-
191
- [More Information Needed]
192
-
193
- ## Model Card Authors [optional]
194
-
195
- [More Information Needed]
196
 
197
  ## Model Card Contact
198
-
199
- [More Information Needed]
 
1
  ---
2
  library_name: transformers
3
+ license: apache-2.0
4
+ language:
5
+ - ja
6
+ - en
7
  ---
8
 
9
+ # Retrieva BERT Model
10
+ The **RetrievaBERT** is the pre-trained Transformer Encoder using Megatron-LM.
11
+ It is designed for use in Japanese.
 
 
12
 
13
  ## Model Details
14
 
15
  ### Model Description
16
 
17
+ The **RetrievaBERT** is the pre-trained Transformer Encoder using Megatron-LM.
 
 
18
 
19
+ It is designed for use in Japanese.
 
 
 
 
 
 
20
 
21
+ This model offers several advanced features compared to traditional BERT models:
22
+ - **PreNorm**: Improved stability during training.
23
+ - **SwiGLU**: Enhanced activation function for better performance.
24
+ - **Grouped-Query Attention (Multi-Query Attention)**: Efficient attention mechanism.
25
+ - **Max Sequence Length**: 2048 tokens, allowing for longer context.
26
+ - **Parameters**: 1.3 billion parameters.
27
+ - **Pre-training Objective**: Only Masked Language Modeling (MLM), not Next Sentence Prediction (NSP).
28
+ - **Token Type IDs**: Not used in this model.
29
 
30
+ ### Model Sources
31
+ - **Developed by:** Retrieva, Inc.
32
+ - **Model type:** Based on MegatronBERT Architecture.
33
+ - **Language(s) (NLP):** Primarily Japanese (optional support for English).
34
+ - **License:** Apache 2.0
35
 
 
 
 
36
 
37
  ## Uses
38
 
39
+ This model can be used as a Masked Language Model (MLM).
40
+ However, it is primarily intended to be fine-tuned on downstream tasks.
41
+ Depending on your use case, follow the appropriate section below.
42
 
43
  ### Direct Use
44
 
45
+ This model is pre-trained using Masked Language Modeling.
46
+ The mask token used is `<MASK|LLM-jp>`.
47
+ Note that you need to set `trust_remote_code` to `True` because RetrievaBERT uses a custom model implementation.
48
+
49
+ Example code for direct use:
 
 
 
 
 
 
 
 
 
 
 
 
 
 
50
 
51
+ ```python
52
+ from transformers import AutoModelForMaskedLM, AutoTokenizer, pipeline
53
 
54
+ model_id = "retrieva-jp/bert-1.3b"
55
+ model = AutoModelForMaskedLM.from_pretrained(model_id, trust_remote_code=True)
56
+ tokenizer = AutoTokenizer.from_pretrained(model_id)
57
+ pipe = pipeline("fill-mask", model=model, tokenizer=tokenizer)
58
 
59
+ text = "こんにちは!私の名前は<MASK|LLM-jp>です!"
60
+ print(pipe(text))
61
+ ```
62
 
63
+ ### Downstream Use
64
 
65
+ RetrievaBERT is compatible with Hugging Face's AutoModels.
66
+ To fine-tune RetrievaBERT for your specific task, use the corresponding AutoModel class.
67
+ For detailed configuration, refer to the config.json file.
68
 
 
 
 
69
 
70
  ## Training Details
71
 
72
  ### Training Data
73
+ The Retrieva BERT model was pre-trained on the reunion of five datasets:
74
+ - [Japanese CommonCrawl Dataset by LLM-jp](https://gitlab.llm-jp.nii.ac.jp/datasets/llm-jp-corpus-v2).
75
+ - [RefinedWeb](https://huggingface.co/datasets/tiiuae/falcon-refinedweb).
76
+ - Chinese Wikipedia dumped on 20240120.
77
+ - Korean Wikipedia dumped on 20240120.
78
+ - [The Stack](https://huggingface.co/datasets/bigcode/the-stack)
79
 
80
+ The model was trained on 180 billion tokens using the above dataset.
 
 
81
 
82
  ### Training Procedure
83
+ The model was trained on 4 to 32 H100 GPUs with a batch size of 1,024.
84
+ We adopted the curriculum learning which is similar to the Sequence Length Warmup and training with the following sequence lengths and number of steps.
85
 
86
+ - The sequence length of 128: 31,000 steps.
87
+ - The sequence length of 256: 219,000 steps.
88
+ - The sequence length of 512: 192,000 steps.
89
+ - The sequence length of 2048: 12,000 steps.
 
 
90
 
91
  #### Training Hyperparameters
92
+ The model was trained on the following hyperparameters.
93
 
94
+ - Learning rate: 1.5e-4.
95
+ - Learning rate decay style: Linear.
96
+ - Learning rate warmup fraction: 0.01
97
+ - Minimum learning rate: 1e-6
98
+ - Floating point expression: BF16
 
 
99
 
100
  ## Evaluation
101
+ We fine-tuned the following models and evaluated them on the [JGLUE](https://github.com/yahoojapan/JGLUE) development set.
102
+ We adjusted the learning rate and training epochs for each model and task in accordance with [the JGLUE paper](https://www.jstage.jst.go.jp/article/jnlp/30/1/30_63/_pdf/-char/ja).
103
 
104
+ | Model | MARC-ja/acc | JSTS/pearson | JSTS/spearman | JNLI/acc | JSQuAD/EM | JSQuAD/F1 | JComQA/acc |
105
+ | :--- |---:|---:|---:|---:|---:|---:|---:|
106
+ | tohoku-nlp/bert-base-japanese-v3 | 0.957 | 0.914 | 0.876 | 0.906 | 0.878 | 0.946 | 0.849 |
107
+ | tohoku-nlp/bert-large-japanese-v2| 0.959 | 0.916 | 0.877 | 0.901 | 0.884 | 0.951 | 0.867 |
108
+ | ku-nlp/deberta-v3-base-japanese    | 0.958 | 0.925 | 0.890 | 0.902 | 0.925 | 0.910 | 0.882 |
109
+ | retrieva-jp/bert-1.3b                        | 0.952 | 0.916 | 0.877 | 0.896 | 0.916 | 0.879 | 0.815 |
 
 
 
 
 
110
 
 
111
 
112
+ ## Technical Specifications
113
 
114
+ ### Model Architectures
115
+ The Retrieva BERT model is based on BERT with the following hyperparameters:
116
 
117
+ - Number of layers: 48
118
+ - Hidden layer size: 1536
119
+ - FFN hidden layer size: 4096
120
+ - Number of attention heads: 24
121
+ - Maximum length of position embeddings: 2048
122
 
123
+ As mentioned earlier, the main differences from the original BERT are:
124
+ - PreNorm: Improved stability during training.
125
+ - SwiGLU: Enhanced activation function for better performance.
126
+ - Grouped-Query Attention (Multi-Query Attention): Efficient attention mechanism.
127
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
128
 
129
  ### Compute Infrastructure
130
 
131
+ [TSUBAME 4](https://www.t4.gsic.titech.ac.jp/en/hardware)
 
 
132
 
133
+ This model is based on results obtained from the TSUBAME deep-learning mini-camp.
134
 
135
  #### Software
136
 
137
+ The model was trained using [Megatron-LM](https://github.com/NVIDIA/Megatron-LM).
 
 
 
 
 
 
 
 
138
 
139
+ ## More Information
140
 
141
+ https://note.com/retrieva/n/n715bea2c2cd1 (in Japanese)
142
 
143
+ ## Model Card Authors
144
 
145
+ Satoru Katsumata, Daisuke Kimura, Jiro Nishitoba
 
 
 
 
 
 
 
 
 
 
146
 
147
  ## Model Card Contact
148