falcon

由技术创新研究院 (TII) 构建的大型语言模型，用于摘要、文本生成和聊天机器人。

7b 40b 180b

59.7K 拉取更新 12个月前

38 个标签

14 个月前更新

14 个月前

7cbd92dfea70 · 84GB

{{- if and .First .System }} {{ .System }} {{- end }} 用户: {{ .Prompt }} 助手:

84B

自述文件

Falcon 是一个由技术创新研究院 (TII) 构建的高性能大型语言模型家族，TII 是阿布扎比政府高级技术研究理事会的一部分，负责监督技术研究。

CLI

ollama run falcon "Why is the sky blue?"

API

curl -X POST https://127.0.0.1:11434/api/generate -d '{
  "model": "falcon",
  "prompt": "Why is the sky blue?"
}'

参数数量

参数数量	推荐内存
70 亿	8GB	查看	`ollama run falcon:7b`
400 亿	32GB	查看	`ollama run falcon:40b`
1800 亿	192GB	查看	`ollama run falcon:180b`

变体


`聊天`	聊天模型在包含多个大型对话数据集的聊天和指令数据集上进行了微调。
`指令`	指令模型会遵循指令，并在 baize 指令数据集上进行了微调。
`文本`	文本模型是基础基础模型，没有针对对话进行微调，最适合用于简单的文本完成。

Falcon 180B

截至 2023 年 9 月，1800 亿参数模型 Falcon 180B 是性能最佳的公开发布的 LLM。它的性能介于 OpenAI 的 GPT 3.5 和 GPT 4 之间。为了运行 Falcon 180B，建议使用至少 192GB 总内存的强大系统。

注意：Falcon 180B 在与较小的同类模型不同的许可证下发布，该许可证在某些情况下限制商业用途。有关更多信息，请参阅模型详细信息和许可证。

更多信息

Falcon is a family of high-performing large language models model built by the Technology Innovation Institute (TII), a research center part of Abu Dhabi government’s advanced technology research council overseeing technology research.

### CLI

```
ollama run falcon "Why is the sky blue?"
```

### API

```
curl -X POST https://127.0.0.1:11434/api/generate -d '{
  "model": "falcon",
  "prompt": "Why is the sky blue?"
}'
```

## Parameter counts
| Parameter Count | Recommended memory |                              |                          |
| --------------- | ------------------ | ---------------------------- | ------------------------ |
| 7 billion       | 8GB               | [View](/library/falcon:7b)   | `ollama run falcon:7b`   |
| 40 billion      | 32GB               | [View](/library/falcon:40b)  | `ollama run falcon:40b`  |
| 180 billion     | 192GB              | [View](/library/falcon:180b) | `ollama run falcon:180b` |

## Variations

|            |                                                                                                                                    |
| ---------- | ---------------------------------------------------------------------------------------------------------------------------------- |
| `chat`     | Chat models are fine-tuned on chat and instruction datasets with a mix of several large-scale conversational datasets.             |
| `instruct` | Instruct models follow instructions and are fine-tuned on the [baize](https://www.google.com/search?q=baize+dataset&oq=baize+data&aqs=chrome.0.0i512j69i57j0i10i15i22i30i625j0i390i650.1387j0j7&sourceid=chrome&ie=UTF-8) instructional dataset.                                                     |
| `text`     | Text models are the base foundation model without any fine-tuning for conversations, and are best used for simple text completion. |

## Falcon 180B

As of September 2023, the 180 billion parameter model, Falcon 180B, is the best-performing openly released LLM. It sits somewhere in between OpenAI's GPT 3.5 and GPT 4. For running Falcon 180B, a powerful system is recommended with at least 192GB of total memory.

> Note: Falcon 180B is released under a different license than its smaller siblings that restricts commercial use under certain conditions. See the [model details](/library/falcon:180b) and license for more information.

## More information

* [TII's website](https://www.tii.ae/)
* [Falcon 180B announcement](https://falconllm.tii.ae)
* [TII on HuggingFace](https://hugging-face.cn/tiiuae)

粘贴、拖放或点击上传图像（.png、.jpeg、.jpg、.svg、.gif）