That is a very old container version (for LLM time), 3 months out of date. My guess is you had it pinned.
If you want to stick with NGC image, pull nvcr.io/nvidia/vllm:26.02-py3 and try again. Or, better yet, clone https://github.com/eugr/spark-vllm-docker and run ./build-and-copy.sh
yes, i pull nvcr.io/nvidia/vllm:26.02-py3 also have this issue, so i try to use old version, not sure why, previous, i can run old version wihtout issue before i factory reset
DGX Spark currently supports version 590 drivers. I am using version 590.48.01 drivers.
±----------------------------------------------------------------------------------------+
| NVIDIA-SMI 590.48.01 Driver Version: 590.48.01 CUDA Version: 13.1 |
±----------------------------------------±-----------------------±---------------------+
| GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|=========================================+========================+======================|
| 0 NVIDIA GB10 On | 0000000F:01:00.0 Off | N/A |
| N/A 73C P0 39W / N/A | Not Supported | 84% Default |
| | | N/A |
±----------------------------------------±-----------------------±---------------------+
±----------------------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=========================================================================================|
| 0 N/A N/A 6337 C …RayWorkerWrapper.ray_call 10873… |
±----------------------------------------------------------------------------------------+
590 drivers have not been rolled into the official update as far as I am able to tell. I’m not surprised things don’t work. Any particular reason for attempting this?