RichardErkhov
commited on
Commit
•
938f7a7
1
Parent(s):
23ee7a7
uploaded readme
Browse files
README.md
ADDED
@@ -0,0 +1,162 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
Quantization made by Richard Erkhov.
|
2 |
+
|
3 |
+
[Github](https://github.com/RichardErkhov)
|
4 |
+
|
5 |
+
[Discord](https://discord.gg/pvy7H8DZMG)
|
6 |
+
|
7 |
+
[Request more models](https://github.com/RichardErkhov/quant_request)
|
8 |
+
|
9 |
+
|
10 |
+
Taiwan-LLM-7B-v2.0-chat - GGUF
|
11 |
+
- Model creator: https://huggingface.co/yentinglin/
|
12 |
+
- Original model: https://huggingface.co/yentinglin/Taiwan-LLM-7B-v2.0-chat/
|
13 |
+
|
14 |
+
|
15 |
+
| Name | Quant method | Size |
|
16 |
+
| ---- | ---- | ---- |
|
17 |
+
| [Taiwan-LLM-7B-v2.0-chat.Q2_K.gguf](https://huggingface.co/RichardErkhov/yentinglin_-_Taiwan-LLM-7B-v2.0-chat-gguf/blob/main/Taiwan-LLM-7B-v2.0-chat.Q2_K.gguf) | Q2_K | 2.36GB |
|
18 |
+
| [Taiwan-LLM-7B-v2.0-chat.IQ3_XS.gguf](https://huggingface.co/RichardErkhov/yentinglin_-_Taiwan-LLM-7B-v2.0-chat-gguf/blob/main/Taiwan-LLM-7B-v2.0-chat.IQ3_XS.gguf) | IQ3_XS | 2.6GB |
|
19 |
+
| [Taiwan-LLM-7B-v2.0-chat.IQ3_S.gguf](https://huggingface.co/RichardErkhov/yentinglin_-_Taiwan-LLM-7B-v2.0-chat-gguf/blob/main/Taiwan-LLM-7B-v2.0-chat.IQ3_S.gguf) | IQ3_S | 2.75GB |
|
20 |
+
| [Taiwan-LLM-7B-v2.0-chat.Q3_K_S.gguf](https://huggingface.co/RichardErkhov/yentinglin_-_Taiwan-LLM-7B-v2.0-chat-gguf/blob/main/Taiwan-LLM-7B-v2.0-chat.Q3_K_S.gguf) | Q3_K_S | 2.75GB |
|
21 |
+
| [Taiwan-LLM-7B-v2.0-chat.IQ3_M.gguf](https://huggingface.co/RichardErkhov/yentinglin_-_Taiwan-LLM-7B-v2.0-chat-gguf/blob/main/Taiwan-LLM-7B-v2.0-chat.IQ3_M.gguf) | IQ3_M | 2.9GB |
|
22 |
+
| [Taiwan-LLM-7B-v2.0-chat.Q3_K.gguf](https://huggingface.co/RichardErkhov/yentinglin_-_Taiwan-LLM-7B-v2.0-chat-gguf/blob/main/Taiwan-LLM-7B-v2.0-chat.Q3_K.gguf) | Q3_K | 3.07GB |
|
23 |
+
| [Taiwan-LLM-7B-v2.0-chat.Q3_K_M.gguf](https://huggingface.co/RichardErkhov/yentinglin_-_Taiwan-LLM-7B-v2.0-chat-gguf/blob/main/Taiwan-LLM-7B-v2.0-chat.Q3_K_M.gguf) | Q3_K_M | 3.07GB |
|
24 |
+
| [Taiwan-LLM-7B-v2.0-chat.Q3_K_L.gguf](https://huggingface.co/RichardErkhov/yentinglin_-_Taiwan-LLM-7B-v2.0-chat-gguf/blob/main/Taiwan-LLM-7B-v2.0-chat.Q3_K_L.gguf) | Q3_K_L | 3.35GB |
|
25 |
+
| [Taiwan-LLM-7B-v2.0-chat.IQ4_XS.gguf](https://huggingface.co/RichardErkhov/yentinglin_-_Taiwan-LLM-7B-v2.0-chat-gguf/blob/main/Taiwan-LLM-7B-v2.0-chat.IQ4_XS.gguf) | IQ4_XS | 3.4GB |
|
26 |
+
| [Taiwan-LLM-7B-v2.0-chat.Q4_0.gguf](https://huggingface.co/RichardErkhov/yentinglin_-_Taiwan-LLM-7B-v2.0-chat-gguf/blob/main/Taiwan-LLM-7B-v2.0-chat.Q4_0.gguf) | Q4_0 | 3.56GB |
|
27 |
+
| [Taiwan-LLM-7B-v2.0-chat.IQ4_NL.gguf](https://huggingface.co/RichardErkhov/yentinglin_-_Taiwan-LLM-7B-v2.0-chat-gguf/blob/main/Taiwan-LLM-7B-v2.0-chat.IQ4_NL.gguf) | IQ4_NL | 3.58GB |
|
28 |
+
| [Taiwan-LLM-7B-v2.0-chat.Q4_K_S.gguf](https://huggingface.co/RichardErkhov/yentinglin_-_Taiwan-LLM-7B-v2.0-chat-gguf/blob/main/Taiwan-LLM-7B-v2.0-chat.Q4_K_S.gguf) | Q4_K_S | 3.59GB |
|
29 |
+
| [Taiwan-LLM-7B-v2.0-chat.Q4_K.gguf](https://huggingface.co/RichardErkhov/yentinglin_-_Taiwan-LLM-7B-v2.0-chat-gguf/blob/main/Taiwan-LLM-7B-v2.0-chat.Q4_K.gguf) | Q4_K | 3.8GB |
|
30 |
+
| [Taiwan-LLM-7B-v2.0-chat.Q4_K_M.gguf](https://huggingface.co/RichardErkhov/yentinglin_-_Taiwan-LLM-7B-v2.0-chat-gguf/blob/main/Taiwan-LLM-7B-v2.0-chat.Q4_K_M.gguf) | Q4_K_M | 3.8GB |
|
31 |
+
| [Taiwan-LLM-7B-v2.0-chat.Q4_1.gguf](https://huggingface.co/RichardErkhov/yentinglin_-_Taiwan-LLM-7B-v2.0-chat-gguf/blob/main/Taiwan-LLM-7B-v2.0-chat.Q4_1.gguf) | Q4_1 | 3.95GB |
|
32 |
+
| [Taiwan-LLM-7B-v2.0-chat.Q5_0.gguf](https://huggingface.co/RichardErkhov/yentinglin_-_Taiwan-LLM-7B-v2.0-chat-gguf/blob/main/Taiwan-LLM-7B-v2.0-chat.Q5_0.gguf) | Q5_0 | 4.33GB |
|
33 |
+
| [Taiwan-LLM-7B-v2.0-chat.Q5_K_S.gguf](https://huggingface.co/RichardErkhov/yentinglin_-_Taiwan-LLM-7B-v2.0-chat-gguf/blob/main/Taiwan-LLM-7B-v2.0-chat.Q5_K_S.gguf) | Q5_K_S | 4.33GB |
|
34 |
+
| [Taiwan-LLM-7B-v2.0-chat.Q5_K.gguf](https://huggingface.co/RichardErkhov/yentinglin_-_Taiwan-LLM-7B-v2.0-chat-gguf/blob/main/Taiwan-LLM-7B-v2.0-chat.Q5_K.gguf) | Q5_K | 4.45GB |
|
35 |
+
| [Taiwan-LLM-7B-v2.0-chat.Q5_K_M.gguf](https://huggingface.co/RichardErkhov/yentinglin_-_Taiwan-LLM-7B-v2.0-chat-gguf/blob/main/Taiwan-LLM-7B-v2.0-chat.Q5_K_M.gguf) | Q5_K_M | 4.45GB |
|
36 |
+
| [Taiwan-LLM-7B-v2.0-chat.Q5_1.gguf](https://huggingface.co/RichardErkhov/yentinglin_-_Taiwan-LLM-7B-v2.0-chat-gguf/blob/main/Taiwan-LLM-7B-v2.0-chat.Q5_1.gguf) | Q5_1 | 4.72GB |
|
37 |
+
| [Taiwan-LLM-7B-v2.0-chat.Q6_K.gguf](https://huggingface.co/RichardErkhov/yentinglin_-_Taiwan-LLM-7B-v2.0-chat-gguf/blob/main/Taiwan-LLM-7B-v2.0-chat.Q6_K.gguf) | Q6_K | 5.15GB |
|
38 |
+
| [Taiwan-LLM-7B-v2.0-chat.Q8_0.gguf](https://huggingface.co/RichardErkhov/yentinglin_-_Taiwan-LLM-7B-v2.0-chat-gguf/blob/main/Taiwan-LLM-7B-v2.0-chat.Q8_0.gguf) | Q8_0 | 6.67GB |
|
39 |
+
|
40 |
+
|
41 |
+
|
42 |
+
|
43 |
+
Original model description:
|
44 |
+
|
45 |
+
---
|
46 |
+
# For reference on model card metadata, see the spec: https://github.com/huggingface/hub-docs/blob/main/modelcard.md?plain=1
|
47 |
+
# Doc / guide: https://huggingface.co/docs/hub/model-cards
|
48 |
+
license: apache-2.0
|
49 |
+
language:
|
50 |
+
- zh
|
51 |
+
widget:
|
52 |
+
- text: >-
|
53 |
+
A chat between a curious user and an artificial intelligence assistant.
|
54 |
+
The assistant gives helpful, detailed, and polite answers to the user's
|
55 |
+
questions. USER: 你好,請問你可以幫我寫一封推薦信嗎? ASSISTANT:
|
56 |
+
library_name: transformers
|
57 |
+
pipeline_tag: text-generation
|
58 |
+
extra_gated_heading: Acknowledge license to accept the repository.
|
59 |
+
extra_gated_prompt: Please contact the author for access.
|
60 |
+
extra_gated_button_content: Acknowledge license 同意以上內容
|
61 |
+
extra_gated_fields:
|
62 |
+
Name: text
|
63 |
+
Mail: text
|
64 |
+
Organization: text
|
65 |
+
Country: text
|
66 |
+
Any utilization of the Taiwan LLM repository mandates the explicit acknowledgment and attribution to the original author: checkbox
|
67 |
+
使用Taiwan LLM必須明確地承認和歸功於優必達株式會社 Ubitus 以及原始作者: checkbox
|
68 |
+
---
|
69 |
+
<img src="https://cdn-uploads.huggingface.co/production/uploads/5df9c78eda6d0311fd3d541f/CmusIT5OlSXvFrbTJ7l-C.png" alt="Taiwan LLM Logo" width="800" style="margin-left:'auto' margin-right:'auto' display:'block'"/>
|
70 |
+
|
71 |
+
# 🌟 Checkout [Taiwan-LLM Demo Chat-UI](http://www.twllm.com) 🌟
|
72 |
+
|
73 |
+
# Model Card for Taiwan LLM 7B v2.0 chat
|
74 |
+
|
75 |
+
Taiwan LLM is an advanced language model tailored for Traditional Chinese, focusing on the linguistic and cultural contexts of Taiwan.
|
76 |
+
Developed from a large base model, it's enriched with diverse Taiwanese textual sources and refined through Supervised Fine-Tuning.
|
77 |
+
This model excels in language understanding and generation, aligning closely with Taiwan's cultural nuances.
|
78 |
+
It demonstrates improved performance on various benchmarks like TC-Eval, showcasing its contextual comprehension and cultural relevance.
|
79 |
+
For detailed insights into Taiwan LLM's development and features, refer to our [technical report](https://github.com/MiuLab/Taiwan-LLaMa/blob/main/twllm_paper.pdf).
|
80 |
+
|
81 |
+
|
82 |
+
## Model description
|
83 |
+
|
84 |
+
- **Model type:** A 7B parameter GPT-like model fine-tuned on a mix of publicly available, synthetic datasets.
|
85 |
+
- **Language(s) (NLP):** Primarily Traditional Chinese (zh-tw)
|
86 |
+
- **Finetuned from model:** [yentinglin/Taiwan-LLM-7B-v2.0-base](https://huggingface.co/yentinglin/yentinglin/Taiwan-LLM-7B-v2.0-base)
|
87 |
+
|
88 |
+
### Model Sources
|
89 |
+
|
90 |
+
<!-- Provide the basic links for the model. -->
|
91 |
+
|
92 |
+
- **Repository:** https://github.com/MiuLab/Taiwan-LLaMa
|
93 |
+
- **Demo:** https://twllm.com/
|
94 |
+
|
95 |
+
## Performance
|
96 |
+
|
97 |
+
|
98 |
+
![image/png](https://cdn-uploads.huggingface.co/production/uploads/5df9c78eda6d0311fd3d541f/HTwIzw6RDha2-PhuWqSuI.png)
|
99 |
+
|
100 |
+
## Intended uses
|
101 |
+
|
102 |
+
Here's how you can run the model using the `pipeline()` function from 🤗 Transformers:
|
103 |
+
|
104 |
+
```python
|
105 |
+
# pip install transformers>=4.34
|
106 |
+
# pip install accelerate
|
107 |
+
|
108 |
+
import torch
|
109 |
+
from transformers import pipeline
|
110 |
+
|
111 |
+
pipe = pipeline("text-generation", model="yentinglin/Taiwan-LLM-7B-v2.0-chat", torch_dtype=torch.bfloat16, device_map="auto")
|
112 |
+
|
113 |
+
# We use the tokenizer's chat template to format each message - see https://huggingface.co/docs/transformers/main/en/chat_templating
|
114 |
+
messages = [
|
115 |
+
{
|
116 |
+
"role": "system",
|
117 |
+
"content": "你是一個人工智慧助理",
|
118 |
+
},
|
119 |
+
{"role": "user", "content": "東北季風如何影響台灣氣候?"},
|
120 |
+
]
|
121 |
+
prompt = pipe.tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
|
122 |
+
outputs = pipe(prompt, max_new_tokens=256, do_sample=True, temperature=0.7, top_k=50, top_p=0.95)
|
123 |
+
print(outputs[0]["generated_text"])
|
124 |
+
```
|
125 |
+
|
126 |
+
### Training hyperparameters
|
127 |
+
|
128 |
+
![image/png](https://cdn-uploads.huggingface.co/production/uploads/5df9c78eda6d0311fd3d541f/MdvHwdUvH-c926qyRAw7K.png)
|
129 |
+
|
130 |
+
![image/png](https://cdn-uploads.huggingface.co/production/uploads/5df9c78eda6d0311fd3d541f/kKpkvxDzOEyiAoTqmzRYO.png)
|
131 |
+
|
132 |
+
|
133 |
+
![image/png](https://cdn-uploads.huggingface.co/production/uploads/5df9c78eda6d0311fd3d541f/FsnlJ_fkRxf7fn5RKZnjE.png)
|
134 |
+
|
135 |
+
The following hyperparameters were used during training:
|
136 |
+
- learning_rate: 5e-05
|
137 |
+
- distributed_type: multi-GPU
|
138 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
139 |
+
- lr_scheduler_type: cosine
|
140 |
+
- lr_scheduler_warmup_ratio: 0.03
|
141 |
+
- num_epochs: 5.0
|
142 |
+
|
143 |
+
## Citation
|
144 |
+
|
145 |
+
If you find Taiwan LLM is useful in your work, please cite it with:
|
146 |
+
|
147 |
+
```
|
148 |
+
@misc{lin2023taiwan,
|
149 |
+
title={Taiwan LLM: Bridging the Linguistic Divide with a Culturally Aligned Language Model},
|
150 |
+
author={Yen-Ting Lin and Yun-Nung Chen},
|
151 |
+
year={2023},
|
152 |
+
eprint={2311.17487},
|
153 |
+
archivePrefix={arXiv},
|
154 |
+
primaryClass={cs.CL}
|
155 |
+
}
|
156 |
+
```
|
157 |
+
|
158 |
+
# Acknowledgement
|
159 |
+
|
160 |
+
Taiwan LLM v2 is conducted in collaboration with [Ubitus K.K.](http://ubitus.net). Ubitus provides valuable compute resources for the project.
|
161 |
+
|
162 |
+
|