Nicholai/jan
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Change env Dockerfile.gpu and update README (#1963)

Browse Source 
Co-authored-by: Hien To <tominhhien97@gmail.com>
This commit is contained in:
hiento09 2024-02-07 21:53:47 +07:00 committed by GitHub
parent 3b0fb4d209
commit 8feaf24186
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
3 changed files with 19 additions and 11 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
Dockerfile.gpu
View File

@ -2,7 +2,7 @@
# Run nvidia-smi to check the CUDA version and the corresponding driver version
# Then update the base image to the appropriate CUDA version refer https://catalog.ngc.nvidia.com/orgs/nvidia/containers/cuda/tags
FROM nvidia/cuda:12.2.2-runtime-ubuntu22.04 AS base
FROM nvidia/cuda:12.2.0-runtime-ubuntu22.04 AS base
# 1. Install dependencies only when needed
FROM base AS builder
@ -73,7 +73,7 @@ RUN npm install -g serve@latest
EXPOSE 1337 3000 3928
ENV LD_LIBRARY_PATH=/usr/local/cuda-12.0/targets/x86_64-linux/lib:/usr/local/cuda-12.0/compat${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}}
ENV LD_LIBRARY_PATH=/usr/local/cuda/targets/x86_64-linux/lib:/usr/local/cuda-12.0/compat${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}}
ENV JAN_API_HOST 0.0.0.0
ENV JAN_API_PORT 1337

9
README.md
View File

@ -236,7 +236,7 @@ This will build the app MacOS m1/m2 for production (with code signing already do
- **Option 1**: Run Jan in CPU mode
```bash
docker compose --profile cpu up
docker compose --profile cpu up -d
```
- **Option 2**: Run Jan in GPU mode
@ -274,18 +274,19 @@ This will build the app MacOS m1/m2 for production (with code signing already do
|=======================================================================================|
```
- **Step 2**: Go to https://catalog.ngc.nvidia.com/orgs/nvidia/containers/cuda/tags and find the lates minor version of image tag that matches the cuda version from the output of `nvidia-smi` (e.g. 12.1 -> 12.1.0)
- **Step 2**: Go to https://catalog.ngc.nvidia.com/orgs/nvidia/containers/cuda/tags and find the smallest minor version of image tag that matches the cuda version from the output of `nvidia-smi` (e.g. 12.1 -> 12.1.0)
- **Step 3**: Update the `Dockerfile.gpu` line number 5 with the latest minor version of the image tag from step 2 (e.g. change `FROM nvidia/cuda:12.2.2-runtime-ubuntu22.04 AS base` to `FROM nvidia/cuda:12.1.0-runtime-ubuntu22.04 AS base`)
- **Step 3**: Update the `Dockerfile.gpu` line number 5 with the latest minor version of the image tag from step 2 (e.g. change `FROM nvidia/cuda:12.2.0-runtime-ubuntu22.04 AS base` to `FROM nvidia/cuda:12.1.0-runtime-ubuntu22.04 AS base`)
- **Step 4**: Run command to start Jan in GPU mode
```bash
# GPU mode
docker compose --profile gpu up
docker compose --profile gpu up -d
```
This will start the web server and you can access Jan at `http://localhost:3000`.
> Note: Currently, Docker mode is only work for development and localhost, production is not supported yet. RAG feature is not supported in Docker mode yet.
## Acknowledgements

17
docker-compose.yml
View File

@ -1,6 +1,9 @@
# Docker Compose file for setting up Minio, createbuckets, app_cpu, and app_gpu services
version: '3.7'
services:
# Minio service for object storage
minio:
image: minio/minio
volumes:
@ -9,6 +12,7 @@ services:
- "9000:9000"
- "9001:9001"
environment:
# Set the root user and password for Minio
MINIO_ROOT_USER: minioadmin # This acts as AWS_ACCESS_KEY
MINIO_ROOT_PASSWORD: minioadmin # This acts as AWS_SECRET_ACCESS_KEY
command: server --console-address ":9001" /data
@ -22,6 +26,7 @@ services:
vpcbr:
ipv4_address: 10.5.0.2
# createbuckets service to create a bucket and set its policy
createbuckets:
image: minio/mc
depends_on:
@ -36,7 +41,7 @@ services:
networks:
vpcbr:
# app_cpu service for running the CPU version of the application
app_cpu:
image: jan:latest
volumes:
@ -45,6 +50,7 @@ services:
context: .
dockerfile: Dockerfile
environment:
# Set the AWS access key, secret access key, bucket name, endpoint, and region for app_cpu
AWS_ACCESS_KEY_ID: minioadmin
AWS_SECRET_ACCESS_KEY: minioadmin
S3_BUCKET_NAME: mybucket
@ -60,8 +66,8 @@ services:
networks:
vpcbr:
ipv4_address: 10.5.0.3
# app_gpu service for running the GPU version of the application
app_gpu:
deploy:
resources:
@ -78,12 +84,12 @@ services:
dockerfile: Dockerfile.gpu
restart: always
environment:
# Set the AWS access key, secret access key, bucket name, endpoint, and region for app_gpu
AWS_ACCESS_KEY_ID: minioadmin
AWS_SECRET_ACCESS_KEY: minioadmin
S3_BUCKET_NAME: mybucket
AWS_ENDPOINT: http://10.5.0.2:9000
AWS_REGION: us-east-1
profiles:
- gpu
ports:
@ -106,5 +112,6 @@ networks:
- subnet: 10.5.0.0/16
gateway: 10.5.0.1
# docker compose --profile cpu up
# docker compose --profile gpu up
# Usage:
# - Run 'docker-compose --profile cpu up -d' to start the app_cpu service
# - Run 'docker-compose --profile gpu up -d' to start the app_gpu service