Hi everyone,
I’m trying to deploy VSS on my DGX Spark, following the guide available here: NVIDIA DGX Spark — Video Search and Summarization Agent
After running all commands in the Fully local Single GPU docker deployment guide , the docker compose up never reaches the final stage of “VIA Server loaded”: a couple seconds after “Starting VIA server in release mode”, I see the “ImportError: libnvrm_mem.so: cannot open shared object file: No such file or directory”, and shortly after, the last via-server log entry: “ProcessGroupNCCL.cpp:1538] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources”
How can I install the set of libnvrm libraries?
This issue got resolved by adding
export IS_SBSA=1
to the .env file.
Subsequently, I receive the following exception:
via-server-1 | 2025-10-23 16:55:07 | ERROR | gradio_web_server | Error loading default prompts: 'functions'
via-server-1 | 2025-10-23 16:55:07 | ERROR | stderr | Traceback (most recent call last):
via-server-1 | 2025-10-23 16:55:07 | ERROR | stderr | File "/opt/nvidia/via/via-engine/via_demo_client.py", line 204, in <module>
via-server-1 | 2025-10-23 16:55:07 | ERROR | stderr | demo = build_demo(args)
via-server-1 | 2025-10-23 16:55:07 | ERROR | stderr | ^^^^^^^^^^^^^^^^
via-server-1 | 2025-10-23 16:55:07 | ERROR | stderr | File "/opt/nvidia/via/via-engine/via_demo_client.py", line 159, in build_demo
via-server-1 | 2025-10-23 16:55:07 | ERROR | stderr | build_summarization(args, appConfig, logger)
via-server-1 | 2025-10-23 16:55:07 | ERROR | stderr | File "/opt/nvidia/via/via-engine/client/summarization.py", line 883, in build_summarization
via-server-1 | 2025-10-23 16:55:07 | ERROR | stderr | (
via-server-1 | 2025-10-23 16:55:07 | ERROR | stderr | ValueError: too many values to unpack (expected 3)
via-server-1 | 2025-10-23 16:55:07 | ERROR | gradio_web_server | Error loading default prompts: 'functions'
via-server-1 | 2025-10-23 16:55:07 | ERROR | stderr | Traceback (most recent call last):
via-server-1 | 2025-10-23 16:55:07 | ERROR | stderr | File "/opt/nvidia/via/via-engine/via_demo_client.py", line 204, in <module>
via-server-1 | 2025-10-23 16:55:07 | ERROR | stderr | demo = build_demo(args)
via-server-1 | 2025-10-23 16:55:07 | ERROR | stderr | ^^^^^^^^^^^^^^^^
via-server-1 | 2025-10-23 16:55:07 | ERROR | stderr | File "/opt/nvidia/via/via-engine/via_demo_client.py", line 159, in build_demo
via-server-1 | 2025-10-23 16:55:07 | ERROR | stderr | build_summarization(args, appConfig, logger)
via-server-1 | 2025-10-23 16:55:07 | ERROR | stderr | File "/opt/nvidia/via/via-engine/client/summarization.py", line 883, in build_summarization
via-server-1 | 2025-10-23 16:55:07 | ERROR | stderr | (
via-server-1 | 2025-10-23 16:55:07 | ERROR | stderr | ValueError: too many values to unpack (expected 3)
via-server-1 | 2025-10-23 16:55:08,934 INFO Initializing Context Manager Handler no.: 5
via-server-1 | 2025-10-23 16:55:08,935 ERROR Error in configuring context manager handler: 'functions'
via-server-1 | 2025-10-23 16:55:08,935 ERROR Traceback (most recent call last):
via-server-1 | File "/usr/local/lib/python3.12/dist-packages/vss_ctx_rag/context_manager/context_manager_handler.py", line 99, in aconfigure
via-server-1 | for function_name in config["functions"]:
via-server-1 | ~~~~~~^^^^^^^^^^^^^
via-server-1 | KeyError: 'functions'
Any help would be appreciated.
Please follow our playbook for how to deploy VSS on DGX Spark: Try NVIDIA NIM APIs
system
Closed
November 17, 2025, 7:33pm
4
This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.