71 lines
2.6 KiB
Markdown
71 lines
2.6 KiB
Markdown
|
||
# License
|
||
|
||
License:
|
||
eyJhbGciOiJSUzI1NiIsImtpZCI6InUzaHgyQjQyQWhEUXM1M0xQY09yNnZhaTdoSlduYnF1bTRZTlZWd1VwSWM9In0.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.VADnc0PExWhGDAxMIbu0vlmPN423B398of4HFl6BMJ1vqGA9H1ESElOZfk0VQ0YnYgwZc_CZF9k0HRyfCBgRhtRKyB1PpGnaKT_kKNVQryykWRpNhnpDqhmTa-wfTUBXNxhu1ktNPKBFNaEbaYuPsLN_aXPGW0dDwp6coGnGEXEqdRmuvekE6ytu7t6IA6flYs35WqCojvvjAmfBdovo2zPTfmlqKeaz7GPrApMo9JBpmb1a6bZEjCoRhhUx_k-v2rbvE3hd9ix9_UMZ6siJ5IKtNuXy_cprcCXXIFVUMcfTnt78RRXY0jCRMQqWkNq9ZGF0Mgcjsh3ts9xSxPgWnw
|
||
|
||
# License 限制
|
||
Add License to your Local Instance
|
||
Create up to 0 teams
|
||
Create up to 1 users
|
||
Store up to 10 GB of data
|
||
Create up to 2 Registered Models
|
||
|
||
Quickstart
|
||
On a machine with Docker and Python installed, run:
|
||
1 pip install wandb --upgrade
|
||
2 wandb server start
|
||
Generate a free license from the Deployer.
|
||
Add it to your W&B Server's localhost's settings.
|
||
Paste the license in the /system-admin page on your localhost
|
||
|
||
# docker 部署
|
||
|
||
deployment:
|
||
version: "3.8"
|
||
|
||
services:
|
||
wandb:
|
||
image: wandb/local:latest
|
||
container_name: wandb-local
|
||
ports:
|
||
- "8080:8080"
|
||
volumes:
|
||
- wandb_data:/vol
|
||
restart: unless-stopped
|
||
|
||
volumes:
|
||
wandb_data:
|
||
|
||
|
||
|
||
# 连接方式：
|
||
方式 B：环境变量（适合容器/批处理/CI）
|
||
|
||
通过 ray job的runtime_env来设置环境变量
|
||
|
||
export WANDB_BASE_URL=http://<服务器IP或域名>:8080
|
||
export WANDB_API_KEY=<你的API_KEY>
|
||
|
||
|
||
官方文档说明可以用 WANDB_BASE_URL + WANDB_API_KEY 代替 wandb login --host ..
|
||
|
||
|
||
# verl配置：
|
||
在 verl 里打开 wandb（你只需要配 trainer）
|
||
|
||
verl 的配置里，最关键是这三个字段：trainer.logger、trainer.project_name、trainer.experiment_name。文档里也写了 logger 用于 console + tracking（tracking 会初始化 wandb）。
|
||
veRL Documentation
|
||
+1
|
||
|
||
推荐写法（新版本）：
|
||
|
||
trainer:
|
||
logger: ["console", "wandb"]
|
||
project_name: my_project # 用argus的用户名_project ，譬如 alice_project
|
||
experiment_name: exp_001 # 用 task id 作为实验名
|
||
|
||
|
||
|
||
|