Hugging Face's logo

tokyotech-llm
/
Swallow-MX-8x7b-NVE-v0.1

Text Generation
Transformers
Safetensors
English
Japanese
mixtral
text-generation-inference
Model card Files
xet
Community
3
Taishi-N324 commited on
Commit
243e7fc
·
verified ·
1 Parent(s): d7e9800

Upload README.md

Browse files
Files changed (1) hide show
  1. README.md +141 -0
README.md ADDED
@@ -0,0 +1,141 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ language:
3
+ - en
4
+ - ja
5
+ library_name: transformers
6
+ pipeline_tag: text-generation
7
+ tag: moe
8
+ license: apache-2.0
9
+ ---
10
+
11
+ # Swallow-MX
12
+
13
+ Our Swallow-MX model has undergone continuous pre-training from the Mixtral-8x7B-Instruct-v0.1, primarily with the addition of Japanese language data.
14
+
15
+ ![logo](./logo.png)
16
+
17
+ ## Model Details
18
+
19
+ * **Model type**: Please refer to Mixtral technical report for details on the model architecture.
20
+ * **Language(s)**: Japanese English
21
+ * **Tokenizer**: This model utilizes the same tokenizer as employed by Mixtral-8x7B-Instruct-v0.1.
22
+ * **Contact**: swallow[at]nlp.c.titech.ac.jp
23
+
24
+ ## Base Model Performance
25
+
26
+ ### Japanese version
27
+
28
+ |Model|Size|JCommonsenseQA|JEMHopQA|NIILC|JSQuAD|XL-Sum|MGSM|WMT20-en-ja|WMT20-ja-en|
29
+ |---|---|---|---|---|---|---|---|---|---|
30
+ | | |4-shot|4-shot|4-shot|4-shot|1-shot|4-shot|4-shot|4-shot|
31
+ | Llama 2 | 7B | 0.3852 | 0.4240 | 0.3410 | 0.7917 | 0.1905 | 0.0760 | 0.1783 | 0.1738 |
32
+ | Swallow | 7B | 0.4808 | 0.5078 | 0.5968 | 0.8573 | 0.1830 | 0.1240 | 0.2510 | 0.1511 |
33
+ | Swallow-Plus | 7B | 0.5478 | 0.5493 | 0.6030 | 0.8544 | 0.1806 | 0.1360 | 0.2568 | 0.1441 |
34
+ | Swallow-NVE | 7B | 0.5433 | 0.5425 | 0.5729 | 0.8684 | 0.2117 | 0.1200 | 0.2405 | 0.1512 |
35
+ | Llama 2 | 13B | 0.6997 | 0.4415 | 0.4170 | 0.8533 | 0.2139 | 0.1320 | 0.2146 | 0.1982 |
36
+ | Swallow | 13B | 0.7837 | 0.5063 | 0.6398 | 0.9005 | 0.2168 | 0.2040 | 0.2720 | 0.1771 |
37
+ | Swallow-NVE | 13B | 0.7712 | 0.5438 | 0.6351 | 0.9030 | 0.2294 | 0.2120 | 0.2735 | 0.1817 |
38
+ | Llama 2 | 70B | 0.8686 | 0.4656 | 0.5256 | 0.9080 | 0.2361 | 0.3560 | 0.2643 | **0.2398** |
39
+ | Swallow | 70B | 0.9348 | **0.6290** | 0.6960 | 0.9176 | 0.2266 | **0.4840** | **0.3043** | 0.2298 |
40
+ | Swallow-NVE | 70B | **0.9410** | 0.5759 | **0.7024** | **0.9254** | **0.2758** | 0.4720 | 0.3042 | 0.2322 |
41
+ |Mixtral-8x7B-v0.1|8x7B|0.8347|0.5335|0.3549|0.8847|0.2192|0.3120|0.1970|0.1987|
42
+ |Swallow-MX-NVE|8x7B|0.9258|0.5843|0.5687|0.9148|0.2589|0.4360|0.2705|0.2074|
43
+
44
+ Please note that Swallow-MX-NVE is not derived from Mixtral-8x7B-v0.1, but rather underwent continued pre-training from Mixtral-8x7B-Instruct-v0.1.
45
+
46
+ ### English version
47
+
48
+ |Model|Size|OpenBookQA|TriviaQA|HellaSwag|SQuAD2.0|XWINO|GSM8K|
49
+ |---|---|---|---|---|---|---|---|
50
+ | | |8-shot|8-shot|8-shot|8-shot|8-shot|8-shot|
51
+ | Llama 2 | 7B | 0.3580 | 0.6265 | 0.5860 | 0.3207 | 0.9049 | 0.1410 |
52
+ | Swallow | 7B | 0.3180 | 0.4836 | 0.5308 | 0.3125 | 0.8817 | 0.1130 |
53
+ | Swallow-Plus | 7B | 0.3280 | 0.4558 | 0.5259 | 0.3134 | 0.8929 | 0.1061 |
54
+ | Swallow-NVE | 7B | 0.3180 | 0.5079 | 0.5329 | 0.2919 | 0.8817 | 0.0986 |
55
+ | Llama 2 | 13B | 0.3760 | 0.7255 | 0.6148 | 0.3681 | 0.9140 | 0.2403 |
56
+ | Swallow | 13B | 0.3500 | 0.5852 | 0.5660 | 0.3406 | 0.9075 | 0.2039 |
57
+ | Swallow-NVE | 13B | 0.3460 | 0.6025 | 0.5700 | 0.3478 | 0.9006 | 0.1751 |
58
+ | Llama 2 | 70B | **0.4280** | **0.8239** | **0.6742** | 0.3770 | **0.9290** | 0.5284 |
59
+ | Swallow | 70B | 0.4220 | 0.7756 | 0.6458 | 0.3745 | 0.9204 | 0.4867 |
60
+ | Swallow-NVE | 70B | 0.4240 | 0.7817 | 0.6439 | 0.3451 | 0.9256 | 0.4943 |
61
+ |Mixtral-8x7B-v0.1|8x7B|0.3960|0.7989|0.6678|**0.3842**|0.9204|**0.5747**|
62
+ |Swallow-MX-NVE|8x7B|0.3740|0.7847|0.6520|0.3801|0.9170|0.5694|
63
+
64
+ Please note that Swallow-MX-NVE is not derived from Mixtral-8x7B-v0.1, but rather underwent continued pre-training from Mixtral-8x7B-Instruct-v0.1.
65
+
66
+ ## Usage
67
+
68
+ First install additional dependencies in [requirements.txt](./requirements.txt):
69
+
70
+ ```sh
71
+ pip install -r requirements.txt
72
+ ```
73
+
74
+ ### Use the base model
75
+
76
+ ```python
77
+ from transformers import AutoModelForCausalLM, AutoTokenizer
78
+
79
+ model_name = "tokyotech-llm/Swallow-MX-NVE-hf"
80
+ tokenizer = AutoTokenizer.from_pretrained(model_name, torch_dtype=torch.bfloat16, device_map="auto")
81
+
82
+ model = AutoModelForCausalLM.from_pretrained(model_name)
83
+ prompt = "東京工業大学の主なキャンパスは、"
84
+ input_ids = tokenizer.encode(
85
+ prompt,
86
+ add_special_tokens=False,
87
+ return_tensors="pt"
88
+ )
89
+ tokens = model.generate(
90
+ input_ids.to(device=model.device),
91
+ max_new_tokens=128,
92
+ temperature=0.99,
93
+ top_p=0.95,
94
+ do_sample=True,
95
+ )
96
+
97
+ out = tokenizer.decode(tokens[0], skip_special_tokens=True)
98
+ print(out)
99
+ ```
100
+
101
+ ## Training Datasets
102
+
103
+ ### Continual Pre-Training
104
+ The following datasets were used for continual pre-training.
105
+
106
+ - [Algebraic Stack](https://huggingface.co/datasets/EleutherAI/proof-pile-2)
107
+ - [Japanese Wikipedia](https://dumps.wikimedia.org/other/cirrussearch)
108
+ - [RefinedWeb](https://huggingface.co/datasets/tiiuae/falcon-refinedweb)
109
+ - [Swallow Corpus](https://chokkan.org/temp/tokyotech-llm/swallow-corpus)
110
+ - [The Pile](https://huggingface.co/datasets/EleutherAI/pile)
111
+ - [The Vault](https://github.com/FSoft-AI4Code/TheVault)
112
+
113
+ ## Risks and Limitations
114
+
115
+ The models released here are still in the early stages of our research and development and have not been tuned to ensure outputs align with human intent and safety considerations.
116
+
117
+ ## Acknowledgements
118
+
119
+ We thank Mistral AI for releasing Mixtral-8x7B-Instruct-v0.1 under an open license for others to build on.
120
+
121
+ Our project is supported by the [ABCI Large-scale Language Model Building Support Program](https://abci.ai/en/link/llm_support_program.html) of the National Institute of Advanced Industrial Science and Technology.
122
+
123
+ ## License
124
+
125
+ apache-2.0
126
+
127
+ ## Authors
128
+
129
+ Here are the team members:
130
+ - From [Okazaki Laboratory](https://www.nlp.c.titech.ac.jp/index.en.html), the following members:
131
+ - [Naoaki Okazaki](https://www.chokkan.org/index.ja.html)
132
+ - [Sakae Mizuki](https://s-mizuki-nlp.github.io/)
133
+ - [Hiroki Iida](https://meshidenn.github.io/)
134
+ - [Mengsay Loem](https://loem-ms.github.io/)
135
+ - [Shota Hirai](https://huggingface.co/Kotemo428)
136
+ - [Kakeru Hattori](https://aya-se.vercel.app/)
137
+ - [Masanari Ohi](https://twitter.com/stjohn2007)
138
+ - From [YOKOTA Laboratory](https://www.rio.gsic.titech.ac.jp/en/index.html), the following members:
139
+ - [Rio Yokota](https://twitter.com/rioyokota)
140
+ - [Kazuki Fujii](https://twitter.com/okoge_kaz)
141
+ - [Taishi Nakamura](https://twitter.com/Setuna7777_2)