mit-han-lab/Llama-3-8B-Instruct-Gradient-1048k-w8a8-per-channel-kv8-per-tensor Updated Feb 14, 2025 • 2