llm-course-hw1 / config.json
spidogan's picture
Added small model with RoPE and MHLA
b600607 verified
{
"dropout": 0.1,
"hidden_dim": 768,
"intermediate_dim": 3072,
"max_seq_len": 128,
"n_head": 12,
"n_kv_head": 12,
"n_layer": 12,
"vocab_size": 1024
}