Skip to content

add support for NVIDIA L20, H20, and AWS/Alibaba Cloud GPU instances #775

Description

@PYTHON01100100

Description

The current hardware selection dashboard is missing key GPU configurations that are heavily used in cloud production environments. We need to expand our supported hardware configuration matrix to include regional market-specific GPUs and core instance profiles from major cloud providers (AWS and Alibaba Cloud).

Suggested Additions

1. Bare Metal / Bare GPU Selection

  • NVIDIA L20: Add configuration options for L20 1×48GB up to L20 8×48GB.
  • NVIDIA H20: Add configuration options for H20 1×96GB up to H20 8×96GB.

2. Cloud Instance Mapping

Amazon Web Services (AWS)
  • P5e Family: NVIDIA H200 (Up to 8× H200, 141GB HBM3e)
  • P5 Family: NVIDIA H100 (Up to 8× H100, 80GB HBM3)
  • P4de / P4d Family: NVIDIA A100 (Up to 8× A100, 80GB/40GB)
  • G6e Family: NVIDIA L40S (Up to 8× L40S, 48GB GDDR6)
  • G6 Family: NVIDIA L4 (Up to 8× L4, 24GB GDDR6)
Alibaba Cloud (ECS)
  • gn8is Series: NVIDIA L20 (48GB VRAM)
  • gn7i / gn7x Series: NVIDIA A10 / T4
  • gn7e / sgn7i Series: NVIDIA A100 clusters

Motivation

Expanding this list ensures developers deploying models via SGLang in AWS or Alibaba Cloud regions can dynamically generate their CLI execution variables directly from the UI without manual parameter math.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions