ollama - 💡(How to fix) Fix Gemma 4 (26b, 31b) crashes with segfault on DGX Spark (ARM64 + Blackwell GB10) [4 comments, 3 participants]

Error from logs: model failed to load, this may be due to resource limitations or an internal error

TL;DR

The issue may be resolved by investigating resource limitations or internal errors specific to the Gemma 4 models on the Ubuntu ARM64 environment.

Guidance

• Verify the system's resource utilization when loading the Gemma 4 models to identify potential bottlenecks.
• Compare the loading process of the working models (Nemotron 3 Super 120B, Qwen 3.5 35B, Qwen 2.5 72B) with the Gemma 4 models to identify any differences in resource allocation or handling.
• Check the journalctl logs for the full register dump to gain more insights into the cause of the segfault.
• Consider testing the Gemma 4 models on a different environment, like the MacBook Pro M3 Max, to see if the issue is specific to the Ubuntu ARM64 setup.

Notes

The issue seems to be specific to the Gemma 4 models on the Ubuntu ARM64 environment, and more investigation is needed to determine the root cause.

Recommendation

Apply workaround: Try to allocate more resources or adjust the loading process for the Gemma 4 models to see if it resolves the issue, as the models work fine on a different hardware setup.