From c463d58277b00820a6af8907e3a26cb924ca5204 Mon Sep 17 00:00:00 2001 From: Daniel Han Date: Sun, 24 May 2026 11:36:36 +0000 Subject: [PATCH] entrypoint.sh: correct driver-floor message (570+ unconditionally on cu128) The earlier message had per-arch driver minimums (525/535/555/570) that came from when each chip first got driver support. That's not how CUDA toolkit floors work -- cu128 imposes 570.26+ on EVERY GPU regardless of arch. Only B300 (sm_103) and DGX Spark (sm_121) need a newer driver (580+), and they ship factory with those drivers anyway. External HF README has the same correction applied in temp/hf_readme.md (updated separately when published). --- docker/entrypoint.sh | 15 ++++++++------- 1 file changed, 8 insertions(+), 7 deletions(-) diff --git a/docker/entrypoint.sh b/docker/entrypoint.sh index 9874ca2c89..7d42407c8b 100755 --- a/docker/entrypoint.sh +++ b/docker/entrypoint.sh @@ -79,15 +79,16 @@ if torch.cuda.is_available(): sys.exit(0) print("ERROR: torch.cuda.is_available() is False despite nvidia-smi working.") print() -print("Most likely the host NVIDIA driver is too old for CUDA 12.8.") -print("Required host driver versions for this image:") -print(" >= 570 RTX 50-series, RTX 6000 Pro Blackwell (sm_120)") -print(" >= 555 B100 / B200 (sm_100)") -print(" >= 535 H100 / H200 (sm_90)") -print(" >= 525 Ada / Ampere (sm_80 / sm_86 / sm_89)") +print("This image bakes in CUDA 12.8, so the host driver MUST be:") +print(" >= 570.26 (toolkit floor for cu128, applies to every GPU)") +print() +print("Two GPUs need an even newer driver because their launch driver was") +print("released after cu128's:") +print(" >= 580 B300 / GB300 (sm_103)") +print(" >= 580 GB10 / DGX Spark (sm_121)") print() print("Check the host (NOT the container) with: nvidia-smi") -print("Then upgrade the driver to match your GPU.") +print("Then upgrade the driver to match.") sys.exit(1) PY