Phi-3-medium-128k-instruct - error due to tensor shape expected 245, got 243

[koureasstavros]

Background & Description

Using LLamaSharp and loading a 128K instruct model like Phi-3-medium-128k-instruct.gguf it throws an error
llama_model_load: error loading model: done_getting_tensors: wrong number of tensors; expected 245, got 243

According to this issue from llama.cpp ggml-org/llama.cpp#7478 this added recently so I suppose if lama.cpp is updated in this library it should be fixed.

API & Usage

Use 128K instruct models

How to implement

Update lama.cpp library

[martindevans]

Could you test if this works with #751? That has a new set of binaries which should include the upstream changes.

[martindevans]

These new binaries have been released, so I'll close this issue now. If it's still a problem please don't hesitate to reopen :)