Model files (vp.gguf)

Read GGUF files, the model format of ggml and llama.cpp.

• read gives the metadata and the tensor table
• tensor gives one tensor as a float32 array, from F32, F16, BF16, Q8_0, Q4_0, Q4_1, Q4_K or Q6_K

Every number equals gguf-py 0.19.0's dequantize bit for bit, in all three engines.

Part of the distribution vipyr; it runs from a checkout of the repository. Planned on PyPI as pip install vipyr.

Functions

vp.gguf.read

Read a GGUF file's metadata and tensor table.

vp.gguf.tensor

One tensor of a GGUF file, dequantized to float32.

Examples in the gallery

• A quantized model file, read in Viv