MiniCPM-V 2.6 Training Tutorial
输入“/”快速插入内容
🔥MiniCPM-V 2.6 Training Tutorial
MiniCPM-V 2.6 Training Tutorial
用户309
用户6987用户36832024年9月6日修改
🙌
Bilibili Accompanying Video: https://www.bilibili.com/video/BV1YT42167mF/?vd_source=cd29f4e20ef69babd26f4f34cc7c8b3f
Suitable Audience: Individuals who have a basic understanding of Linux and can read Python scripts, modify parameters, and handle data.
1.
Get the GitHub code of MiniCPM-V:
代码块
git clone https://github.com/OpenBMB/MiniCPM-V.git
2.
Install dependencies
代码块
cd MiniCPM-V
pip install -r requirements.txt
3.
Process the dataset
1.
Process data integrated in the following form：
a.
The id value cannot be repeated
b.
<image>\n should appear at the beginning of the conversation data in each data set
c.
The address corresponding to "image" needs to contain an image.
d.
The list corresponding to each conversation is a multi-turn conversation, content represents the conversation content, role corresponds to user and represents user input, and role corresponds to assistant and represents model output.
Single image
代码块
[
{ "id": "0",
"conversations": [
{
"content": "<image>\nWho are they?",
"role": "user"
},
{
"content": "They're Kane and Gretzka from Bayern Munich.",
"role": "assistant"
},
{
"content": "What are they doing?",
"role": "user"
},
{
"content": "They are celebrating on the soccer field.",
"role": "assistant"
}
],
"image":
"/root/ld/ld_project/LLaMA-Factory/data/mllm_demo_data/1.jpg"
}
...The above is a single piece of data. Multiple data in the same format can exist in each list.
]
Multi -image
代码块
[
{
"id": "0",
"image": {
"<image_00>": "path/to/image_0.jpg",
"<image_01>": "path/to/image_1.jpg",
"<image_02>": "path/to/image_2.jpg",
"<image_03>": "path/to/image_3.jpg"
},
"conversations": [
{
"role": "user",
"content": "How to create such text-only videos using CapCut?\n<image_00>\n<image_01>\n<image_02>\n<image_03>\n"
},
{
"role": "assistant",
"content": "To create a text-only video as shown in the images, follow these steps in CapCut..."
}
]
}
]