docs: fix the customize engine section
This commit is contained in:
Arista Indrajaya
parent
701dd309bc
commit
3d24af5504
@ -1,185 +1,63 @@
|
|||||||
---
|
---
|
||||||
sidebar_position: 3
|
sidebar_position: 1
|
||||||
---
|
---
|
||||||
|
|
||||||
import Tabs from '@theme/Tabs';
|
import Tabs from '@theme/Tabs';
|
||||||
import TabItem from '@theme/TabItem';
|
import TabItem from '@theme/TabItem';
|
||||||
import janModel from './assets/jan-model-hub.png';
|
|
||||||
|
|
||||||
# Manual Import
|
# Customize Engine Settings
|
||||||
|
|
||||||
:::warning
|
In this guide, we'll walk you through the process of customizing your engine settings by tweaking the `nitro.json` file
|
||||||
|
|
||||||
This is currently under development.
|
1. Navigate to the `App Settings` > `Advanced` > `Open App Directory` > `~/jan/engine` folder.
|
||||||
|
|
||||||
|
<Tabs>
|
||||||
|
<TabItem value="mac" label="MacOS" default>
|
||||||
|
```sh
|
||||||
|
cd ~/jan/engines
|
||||||
|
```
|
||||||
|
</TabItem>
|
||||||
|
<TabItem value="windows" label="Windows" default>
|
||||||
|
```sh
|
||||||
|
C:/Users/<your_user_name>/jan/engines
|
||||||
|
```
|
||||||
|
</TabItem>
|
||||||
|
<TabItem value="linux" label="Linux" default>
|
||||||
|
```sh
|
||||||
|
cd ~/jan/engines
|
||||||
|
```
|
||||||
|
</TabItem>
|
||||||
|
</Tabs>
|
||||||
|
|
||||||
|
2. Modify the `nitro.json` file based on your needs. The default settings are shown below.
|
||||||
|
|
||||||
|
```json title="~/jan/engines/nitro.json"
|
||||||
|
{
|
||||||
|
"ctx_len": 2048,
|
||||||
|
"ngl": 100,
|
||||||
|
"cpu_threads": 1,
|
||||||
|
"cont_batching": false,
|
||||||
|
"embedding": false
|
||||||
|
}
|
||||||
|
```
|
||||||
|
|
||||||
|
The table below describes the parameters in the `nitro.json` file.
|
||||||
|
|
||||||
|
| Parameter | Type | Description |
|
||||||
|
| --------- | ---- | ----------- |
|
||||||
|
| `ctx_len` | **Integer** | Typically set at `2048`, `ctx_len` provides ample context for model operations like `GPT-3.5`. (*Maximum*: `4096`, *Minimum*: `1`) |
|
||||||
|
| `ngl` | **Integer** | Defaulted at `100`, `ngl` determines GPU layer usage. |
|
||||||
|
| `cpu_threads` | **Integer** | Determines CPU inference threads, limited by hardware and OS. (*Maximum* determined by system) |
|
||||||
|
| `cont_batching` | **Integer** | Controls continuous batching, enhancing throughput for LLM inference. |
|
||||||
|
| `embedding` | **Integer** | Enables embedding utilization for tasks like document-enhanced chat in RAG-based applications. |
|
||||||
|
|
||||||
|
:::tip
|
||||||
|
- By default, the value of `ngl` is set to 100, which indicates that it will offload all. If you wish to offload only 50% of the GPU, you can set `ngl` to 15 because most models on Mistral or Llama are around ~ 30 layers.
|
||||||
|
- To utilize the embedding feature, include the JSON parameter `"embedding": true`. It will enable Nitro to process inferences with embedding capabilities. Please refer to the [Embedding in the Nitro documentation](https://nitro.jan.ai/features/embed) for a more detailed explanation.
|
||||||
|
- To utilize the continuous batching feature for boosting throughput and minimizing latency in large language model (LLM) inference, include `cont_batching: true`. For details, please refer to the [Continuous Batching in the Nitro documentation](https://nitro.jan.ai/features/cont-batch).
|
||||||
|
|
||||||
:::
|
:::
|
||||||
|
|
||||||
This section will show you how to perform manual import. In this guide, we are using a GGUF model from [HuggingFace](https://huggingface.co/) and our latest model, [Trinity](https://huggingface.co/janhq/trinity-v1-GGUF), as an example.
|
|
||||||
|
|
||||||
## Newer versions - nightly versions and v0.4.4+
|
|
||||||
|
|
||||||
### 1. Create a Model Folder
|
|
||||||
|
|
||||||
1. Navigate to the `App Settings` > `Advanced` > `Open App Directory` > `~/jan/models` folder.
|
|
||||||
|
|
||||||
<Tabs>
|
|
||||||
<TabItem value="mac" label="MacOS" default>
|
|
||||||
```sh
|
|
||||||
cd ~/jan/models
|
|
||||||
```
|
|
||||||
</TabItem>
|
|
||||||
<TabItem value="windows" label="Windows" default>
|
|
||||||
```sh
|
|
||||||
C:/Users/<your_user_name>/jan/models
|
|
||||||
```
|
|
||||||
</TabItem>
|
|
||||||
<TabItem value="linux" label="Linux" default>
|
|
||||||
```sh
|
|
||||||
cd ~/jan/models
|
|
||||||
```
|
|
||||||
</TabItem>
|
|
||||||
</Tabs>
|
|
||||||
|
|
||||||
2. In the `models` folder, create a folder with the name of the model.
|
|
||||||
|
|
||||||
```sh
|
|
||||||
mkdir trinity-v1-7b
|
|
||||||
```
|
|
||||||
|
|
||||||
### 2. Drag & Drop the Model
|
|
||||||
|
|
||||||
Drag and drop your model binary into this folder, ensuring the `modelname.gguf` is the same name as the folder name, e.g. `models/modelname`.
|
|
||||||
|
|
||||||
### 3. Done!
|
|
||||||
|
|
||||||
If your model doesn't show up in the **Model Selector** in conversations, **restart the app** or contact us via our [Discord community](https://discord.gg/Dt7MxDyNNZ).
|
|
||||||
|
|
||||||
## Older versions - before v0.4.4
|
|
||||||
|
|
||||||
### 1. Create a Model Folder
|
|
||||||
|
|
||||||
1. Navigate to the `App Settings` > `Advanced` > `Open App Directory` > `~/jan/models` folder.
|
|
||||||
|
|
||||||
<Tabs>
|
|
||||||
<TabItem value="mac" label="MacOS" default>
|
|
||||||
```sh
|
|
||||||
cd ~/jan/models
|
|
||||||
```
|
|
||||||
</TabItem>
|
|
||||||
<TabItem value="windows" label="Windows" default>
|
|
||||||
```sh
|
|
||||||
C:/Users/<your_user_name>/jan/models
|
|
||||||
```
|
|
||||||
</TabItem>
|
|
||||||
<TabItem value="linux" label="Linux" default>
|
|
||||||
```sh
|
|
||||||
cd ~/jan/models
|
|
||||||
```
|
|
||||||
</TabItem>
|
|
||||||
</Tabs>
|
|
||||||
|
|
||||||
2. In the `models` folder, create a folder with the name of the model.
|
|
||||||
|
|
||||||
```sh
|
|
||||||
mkdir trinity-v1-7b
|
|
||||||
```
|
|
||||||
|
|
||||||
### 2. Create a Model JSON
|
|
||||||
|
|
||||||
Jan follows a folder-based, [standard model template](https://jan.ai/docs/engineering/models/) called a `model.json` to persist the model configurations on your local filesystem.
|
|
||||||
|
|
||||||
This means that you can easily reconfigure your models, export them, and share your preferences transparently.
|
|
||||||
|
|
||||||
<Tabs>
|
|
||||||
<TabItem value="mac" label="MacOS" default>
|
|
||||||
```sh
|
|
||||||
cd trinity-v1-7b
|
|
||||||
touch model.json
|
|
||||||
```
|
|
||||||
</TabItem>
|
|
||||||
<TabItem value="windows" label="Windows" default>
|
|
||||||
```sh
|
|
||||||
cd trinity-v1-7b
|
|
||||||
echo {} > model.json
|
|
||||||
```
|
|
||||||
</TabItem>
|
|
||||||
<TabItem value="linux" label="Linux" default>
|
|
||||||
```sh
|
|
||||||
cd trinity-v1-7b
|
|
||||||
touch model.json
|
|
||||||
```
|
|
||||||
</TabItem>
|
|
||||||
</Tabs>
|
|
||||||
|
|
||||||
To update `model.json`:
|
|
||||||
|
|
||||||
- Match `id` with folder name.
|
|
||||||
- Ensure GGUF filename matches `id`.
|
|
||||||
- Set `source.url` to direct download link ending in `.gguf`. In HuggingFace, you can find the direct links in the `Files and versions` tab.
|
|
||||||
- Verify that you are using the correct `prompt_template`. This is usually provided in the HuggingFace model's description page.
|
|
||||||
|
|
||||||
```json title="model.json"
|
|
||||||
{
|
|
||||||
"sources": [
|
|
||||||
{
|
|
||||||
"filename": "trinity-v1.Q4_K_M.gguf",
|
|
||||||
"url": "https://huggingface.co/janhq/trinity-v1-GGUF/resolve/main/trinity-v1.Q4_K_M.gguf"
|
|
||||||
}
|
|
||||||
],
|
|
||||||
"id": "trinity-v1-7b",
|
|
||||||
"object": "model",
|
|
||||||
"name": "Trinity-v1 7B Q4",
|
|
||||||
"version": "1.0",
|
|
||||||
"description": "Trinity is an experimental model merge of GreenNodeLM & LeoScorpius using the Slerp method. Recommended for daily assistance purposes.",
|
|
||||||
"format": "gguf",
|
|
||||||
"settings": {
|
|
||||||
"ctx_len": 4096,
|
|
||||||
"prompt_template": "{system_message}\n### Instruction:\n{prompt}\n### Response:",
|
|
||||||
"llama_model_path": "trinity-v1.Q4_K_M.gguf"
|
|
||||||
},
|
|
||||||
"parameters": {
|
|
||||||
"max_tokens": 4096
|
|
||||||
},
|
|
||||||
"metadata": {
|
|
||||||
"author": "Jan",
|
|
||||||
"tags": ["7B", "Merged"],
|
|
||||||
"size": 4370000000
|
|
||||||
},
|
|
||||||
"engine": "nitro"
|
|
||||||
}
|
|
||||||
```
|
|
||||||
#### Regarding `model.json`
|
|
||||||
|
|
||||||
- In `settings`, two crucial values are:
|
|
||||||
- `ctx_len`: Defined based on the model's context size.
|
|
||||||
- `prompt_template`: Defined based on the model's trained template (e.g., ChatML, Alpaca).
|
|
||||||
- To set up the `prompt_template`:
|
|
||||||
1. Visit Hugging Face.
|
|
||||||
2. Locate the model (e.g., [Gemma 7b it](https://huggingface.co/google/gemma-7b-it)).
|
|
||||||
3. Review the text and identify the template.
|
|
||||||
- In `parameters`, consider the following options. The fields in `parameters` are typically general and can be the same across models. An example is provided below:
|
|
||||||
|
|
||||||
```json
|
|
||||||
"parameters":{
|
|
||||||
"temperature": 0.7,
|
|
||||||
"top_p": 0.95,
|
|
||||||
"stream": true,
|
|
||||||
"max_tokens": 4096,
|
|
||||||
"frequency_penalty": 0,
|
|
||||||
"presence_penalty": 0
|
|
||||||
}
|
|
||||||
```
|
|
||||||
|
|
||||||
### 3. Download the Model
|
|
||||||
|
|
||||||
1. Restart Jan and navigate to the Hub.
|
|
||||||
2. Locate your model.
|
|
||||||
3. Click **Download** button to download the model binary.
|
|
||||||
|
|
||||||
<div class="text--center">
|
|
||||||
<img src={janModel} width={800} alt="jan-model-hub" />
|
|
||||||
</div>
|
|
||||||
|
|
||||||
:::info[Assistance and Support]
|
:::info[Assistance and Support]
|
||||||
|
|
||||||
If you have questions, please join our [Discord community](https://discord.gg/Dt7MxDyNNZ) for support, updates, and discussions.
|
If you have questions, please join our [Discord community](https://discord.gg/Dt7MxDyNNZ) for support, updates, and discussions.
|
||||||
|
|||||||
Loading…
x
Reference in New Issue
Block a user