Anemll DSpark vLLM Runtime Base Image for GB10: Specification and Deployment Guide

The Anemll DSpark vLLM runtime uses the Docker base image ghcr.io/anemll/dspark-vllm-gx10:0.1.1 for all GB10 deployments, as configured in the MiaAI-Lab DeepSeek-v4-Flash-DSpark-2x-DGX-Spark repository.

The MiaAI-Lab repository provides containerized inference capabilities optimized for NVIDIA GB10 hardware. Understanding the exact base image specification is critical for reproducible deployments, custom builds, and troubleshooting container orchestration.

GB10 Base Image Specification

Image Registry and Version

The runtime relies on a GPU-optimized container hosted on GitHub Container Registry. According to the repository's README (lines 17-18), the default runtime image is:

  • Repository: ghcr.io/anemll/dspark-vllm-gx10
  • Tag: 0.1.1
  • Full Path: ghcr.io/anemll/dspark-vllm-gx10:0.1.1

This image serves as the foundation for the Dockerfile.gb10-dsv4-dspark build process, which constructs the GB10-specific DSpark variant.

Deploying the Anemll DSpark Base Image

Pulling the Container Locally

To fetch the base image for local development or inspection:

docker pull ghcr.io/anemll/dspark-vllm-gx10:0.1.1

Configuring Docker Compose Services

When orchestrating multi-container deployments, reference the image directly in your service definition:

services:
  dspam
    image: ghcr.io/anemll/dspark-vllm-gx10:0.1.1   # ← base image

    environment:
      - DSPARK_VLLM_IMAGE=ghcr.io/anemll/dspark-vllm-gx10:0.1.1
    # …other configuration…

Environment Variable Configuration

The repository uses the DSPARK_VLLM_IMAGE variable to standardize image references across deployment scripts. Set this in your .env.dspark file:


# .env.dspark

DSPARK_VLLM_IMAGE=ghcr.io/anemll/dspark-vllm-gx10:0.1.1

Source Files and Build Context

Three configuration files in the MiaAI-Lab repository manage the base image specification:

  • README.md — Documents the default image at lines 17-18, establishing ghcr.io/anemll/dspark-vllm-gx10:0.1.1 as the standard runtime target for GB10 hardware.

  • Dockerfile.gb10-dsv4-dspark — The build instruction file that constructs the GB10-optimized DSpark image, inheriting from the specified base image.

  • .env.dspark.example — Template configuration providing the DSPARK_VLLM_IMAGE environment variable, ensuring consistent image resolution across development and production environments.

Summary

  • The standard Anemll DSpark vLLM base image for GB10 deployments is ghcr.io/anemll/dspark-vllm-gx10:0.1.1
  • Configure container resolution using the DSPARK_VLLM_IMAGE environment variable in .env.dspark
  • Reference the image in orchestration files via the image: directive in Docker Compose
  • Extend or customize builds using Dockerfile.gb10-dsv4-dspark while maintaining the base image dependency

Frequently Asked Questions

What is the exact registry path for the Anemll DSpark GB10 base image?

The complete registry path is ghcr.io/anemll/dspark-vllm-gx10:0.1.1. This GitHub Container Registry URL is specified in the repository's README and environment templates as the default runtime for GB10 hardware.

How do I override the default base image in a DSpark deployment?

Set the DSPARK_VLLM_IMAGE environment variable in your .env.dspark file or export it in your shell before running Docker Compose. This variable takes precedence in the repository's launch scripts.

Which Dockerfile builds the GB10-specific DSpark runtime?

The Dockerfile.gb10-dsv4-dspark file in the repository root contains the build instructions for the GB10-optimized variant, using ghcr.io/anemll/dspark-vllm-gx10:0.1.1 as its parent image.

Version 0.1.1 is the current stable release specified in the default configuration files. Pin to this specific tag rather than latest to ensure reproducible deployments across environments.

Have a question about this repo?

These articles cover the highlights, but your codebase questions are specific. Give your agent direct access to the source. Share this with your agent to get started:

Share the following with your agent to get started:
curl -s "https://instagit.com/install.md"

Works with
Claude Codex Cursor VS Code OpenClaw Any MCP Client

Maintain an open-source project? Get it listed too →