naija-petro-GGUF/README.md

license, language, base_model, pipeline_tag, tags

license	language	base_model	pipeline_tag	tags
apache-2.0	en	Shinzmann/naija-petro	text-generation	petroleum-engineering oil-and-gas nigeria qwen3 gguf llama-cpp ollama quantized

Naija-Petro (32B) — GGUF

GGUF quantizations of Shinzmann/naija-petro (the 32B variant) for inference with llama.cpp, Ollama, LM Studio, and compatible runtimes.

See the full model card for training details, intended use, and limitations. For a lighter footprint, consider the 8B GGUF. For Nigeria-specific accuracy, use these weights with the Naija-Petro RAG system.

Available quantizations

File suffix	Method	Notes
Q4_K_M	4-bit (k-quant, medium)	Smallest; recommended default for 32B on limited RAM/VRAM
Q5_K_M	5-bit (k-quant, medium)	Higher quality, larger
Q8_0	8-bit	Near-lossless; largest and slowest

32B GGUF files are large. Q4_K_M is the practical choice for most machines; ensure you have enough RAM/VRAM + disk for the chosen quant.

Usage

Ollama

ollama run hf.co/Shinzmann/naija-petro-GGUF:Q4_K_M

llama.cpp

./llama-cli -hf Shinzmann/naija-petro-GGUF:Q4_K_M \
  -p "What are the screening criteria for CO2 EOR?" -c 4096

License

Apache-2.0 (inherited from Qwen3-32B). Validate outputs with qualified engineers before operational use.