Downgrade GPU base image to CUDA 12.8.1; add benchmark script

CUDA 13.3 requires driver >= 575 but the host only has 570 (error 804).
CUDA 12.8.1 is the highest version supported by driver 570 and works
correctly with the NVIDIA Container Toolkit.

Add scripts/benchmark.py to measure InsightFace + SigLIP latency and
throughput across GPU and CPU modes.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
This commit is contained in:
2026-06-13 22:19:16 +00:00
co-authored by Claude Sonnet 4.6
parent 8d1f5da05a
commit 9f0a78522f
2 changed files with 197 additions and 2 deletions
+2 -2
View File
@@ -1,5 +1,5 @@
# ── Base images ───────────────────────────────────────────────────────────────
# amd64 + gpu: NVIDIA CUDA 13.3 + cuDNN (GPU acceleration via NVIDIA Container Toolkit)
# amd64 + gpu: NVIDIA CUDA 12.8 + cuDNN (GPU acceleration via NVIDIA Container Toolkit)
# amd64 + rocm: Ubuntu 22.04 (AMD GPU via ROCm — pass /dev/kfd and /dev/dri)
# amd64 + intel: Ubuntu 22.04 (Intel Arc / iGPU via OpenVINO — pass /dev/dri)
# amd64 + cpu: Ubuntu 22.04 (CPU-only, ~2 GB smaller image)
@@ -7,7 +7,7 @@
ARG VARIANT=gpu
FROM --platform=$BUILDPLATFORM nvidia/cuda:13.3.0-cudnn-runtime-ubuntu22.04 AS base-amd64-gpu
FROM --platform=$BUILDPLATFORM nvidia/cuda:12.8.1-cudnn-runtime-ubuntu22.04 AS base-amd64-gpu
FROM ubuntu:22.04 AS base-amd64-rocm
FROM ubuntu:22.04 AS base-amd64-intel
FROM ubuntu:22.04 AS base-amd64-cpu