ahoylabs / gguf.js Goto Github PK

Hi! Thanks for making this library!
I tried two models with it, phind-codellama-34b-v2.Q4_K_M.gguf worked fine, but I got an error with another one:

 error: ZodError: [
    {
      "code": "invalid_type",
      "expected": "number",
      "received": "undefined",
      "path": [
        "general",
        "quantization_version"
      ],
      "message": "Required"
    }
  ]

model: orca-mini-v2_7b.ggmlv3.q4_0.bin
converted to GGUF with https://github.com/ggerganov/llama.cpp/blob/master/convert-llama-ggml-to-gguf.py
it was converted before this commit ggerganov/llama.cpp@ea2c85d

P.S. Do you have any plans on providing direct layer/weight data access (R/W?) Or is this library only for metadata?