Edit model card

Meta-Llama-3-8B-Instruct - bnb 4bit

Description

This model is 4bit quantized version of Meta-Llama-3-8B-Instruct using bitsandbytes. It's designed for fine-tuning! The PAD token is set as UNK.

Downloads last month
2
Safetensors
Model size
4.65B params
Tensor type
F32
·
BF16
·
U8
·
Inference API
Input a message to start chatting with leliuga/Meta-Llama-3-8B-Instruct-bnb-4bit.
Inference API (serverless) has been turned off for this model.

Quantized from

Spaces using leliuga/Meta-Llama-3-8B-Instruct-bnb-4bit 2

Collection including leliuga/Meta-Llama-3-8B-Instruct-bnb-4bit