KVInfer / config_llama.json
NOT-OMEGA's picture
Initial commit: Uploading KVInfer Custom C++ Model (Llama 3B)
7e68fe0 verified
{
"n_layer": 28,
"n_head": 24,
"n_kv_head": 8,
"n_embd": 3072,
"n_intermediate": 8192,
"vocab_size": 128256,
"max_seq_len": 2048,
"rope_theta": 500000.0
}