withsecure / llama3-8b-prompt-injection

This is a quantized version of Llama3-8B, fine-tuned to experiment with enhanced resistance to common indirect prompt injection attacks.

8B

2 Pulls Updated 22 hours ago

... /
params
c0aac7c7f00d · 128B
{ "stop": [ "<|start_header_id|>", "<|end_header_id|>", "<|eot_id|>", "<|reserved_special_token" ] }