Where is the config.json?
Can you add the config.json?
Hi,
Apologies for the late reply, thanks for reaching out to us. The GGUF (GPT-NeoX and other models Unify Format) file format is a self-contained, monolithic file. Self-Contained: The GGUF file itself is a container that includes all the model metadata and configuration (like layer count, head count, context length, etc.) in its header, along with the quantized weights.
Unlike the standard Hugging Face Transformers format, which requires separate files like config.json, tokenizer.json, and the weight files, the GGUF format combines everything into a single .gguf file.
To get the core parameters equivalent to what's in a config.json, you should look at the un-quantized base model or the model's documentation.
The architectural and high-level configuration details are defined in the original, un-quantized model repository for this model the base/original model is google/gemma-3-27b-it or your could use the llama-cli tom extract the meta data in formation ./llama.cpp/bin/llama-cli --model ./gemma-3-27b-it-q4_0.gguf --dump-gguf.
Thanks.