rmarnold

Results 2 comments of rmarnold

[Feature] Export Lora Adapters as GGML

@awni, Yes, in llama.cpp you can specify --lora ggml-adapter-model.bin. The issue is that after training, mlx_ml outputs adapter.safetensors, which llama.cpp does not recognize. I know for certain that the supported...

rope_scaling's short_factor field must have length 64, got 48 when using max_seq_length=64 in Phi-4-mini-instruct-unsloth-bnb-4bi

I think Microsoft patched some of required libraries - https://huggingface.co/microsoft/Phi-4-mini-instruct/discussions/1#67c2179f95640f2cec730fd3 transformers==4.49.0 vllm==0.7.3