Model not working in Ollama?
I get this error when I try to run it:
ollama run hf.co/NikolayKozloff/phi-4-Q5_K_S-GGUF
Error: llama runner process has terminated: GGML_ASSERT(hparams.n_swa > 0) failed
Running ollama version is 0.5.4
Whereas the Phi-4 from the Ollama library is working just fine:
ollama run phi4:14b-q4_K_M
pulling manifest
pulling fd7b6731c33c... 100% ▕████████████████████████████████████████████████████████▏ 9.1 GB
pulling 32695b892af8... 100% ▕████████████████████████████████████████████████████████▏ 275 B
pulling fa8235e5b48f... 100% ▕████████████████████████████████████████████████████████▏ 1.1 KB
pulling 45a1c652dddc... 100% ▕████████████████████████████████████████████████████████▏ 82 B
pulling f5d6f49c6477... 100% ▕████████████████████████████████████████████████████████▏ 486 B
verifying sha256 digest
writing manifest
success
See this thread on a similar quantized model for explanation - the problem is Ollama is behind llama.cpp with functionality, it should be remedied in the next day or two then Ollama users will need to update to the latest version for this to work