4-bit quantized weights (q4_0) of TRL-Lib's Stack LLaMa (based on Meta's 7B LLaMa model), to be used with llama.cpp

TRL-Lib's adapter: https://huggingface.co/trl-lib/llama-7b-se-rl-peft

llama.cpp project: https://github.com/ggerganov/llama.cpp

This is the full model weights of Meta's original model with TRL-Lib's adapter already applied, not the LoRA weights. With this in mind, make sure you are complying with Meta's original licensing by applying for access at the following link. https://ai.facebook.com/blog/large-language-model-llama-meta-ai/

Original HF LLaMa model: https://huggingface.co/decapoda-research/llama-7b-hf

The weights bin file should be placed in the models/trl-lib subdirectory of llama.cpp.

This is a ggjt version of the model, the tensors are 32 byte aligned to allow mmap loading.