llms-from-scratch-cn/Model_Architecture_Discussions/llama3/params.json
2024-05-27 15:11:03 +08:00

11 lines
212 B
JSON

{
"dim": 4096,
"n_layers": 32,
"n_heads": 32,
"n_kv_heads": 8,
"vocab_size": 128256,
"multiple_of": 1024,
"ffn_dim_multiplier": 1.3,
"norm_eps": 1e-05,
"rope_theta": 500000.0
}