镜像地址

我们提供了预构建的Docker镜像以便快速开始：

• torch2.6.0 + SGlang0.4.6: roll-registry.cn-hangzhou.cr.aliyuncs.com/roll/pytorch:nvcr-24.05-py3-torch260-sglang046
• torch2.6.0 + vLLM0.8.4: roll-registry.cn-hangzhou.cr.aliyuncs.com/roll/pytorch:nvcr-24.05-py3-torch260-vllm084
• torch2.8.0 + vLLM0.10.2: roll-registry-vpc.cn-hangzhou.cr.aliyuncs.com/roll/pytorch:nvcr-25.06-py3-torch280-vllm0102
• torch2.8.0 + vLLM0.11.0: roll-registry.cn-hangzhou.cr.aliyuncs.com/roll/pytorch:nvcr-25.06-py3-torch280-vllm0110
• torch2.10.0 + vLLM0.16.0rc2.dev502+gade81f17f + megatron-core core_dev_r0.16.0: roll-registry.cn-hangzhou.cr.aliyuncs.com/roll/pytorch:nvcr-25.11-py3-torch2100-mcore0160dev-vllm016dev

您也可以在docker/目录下找到Dockerfiles来构建您自己的镜像。