autoSLM / assets /config.json
Nihal2000's picture
git add
6145f50
raw
history blame contribute delete
462 Bytes
{
"model_name": "Automotive-SLM-Edge-3M",
"d_model": 256,
"n_layer": 4,
"n_head": 4,
"vocab_size": 50257,
"n_positions": 256,
"use_moe": true,
"n_experts": 4,
"expert_capacity": 2,
"moe_intermediate_size": 384,
"router_aux_loss_coef": 0.01,
"rotary_dim": 64,
"rope_base": 10000,
"dropout": 0.05,
"layer_norm_epsilon": 1e-05,
"max_gen_length": 50,
"temperature": 0.8,
"top_p": 0.9,
"top_k": 50,
"repetition_penalty": 1.1
}