Aurora-Proelia-ChatML / config.json
arthu1's picture
Add standard Transformers inference adapter
5caaad2 verified
Raw
History Blame Contribute Delete
919 Bytes
{
"_name_or_path": "North-ML1/Aurora-Proelia-ChatML",
"architectures": ["AuroraForCausalLM"],
"auto_map": {
"AutoConfig": "configuration_aurora.AuroraHFConfig",
"AutoModelForCausalLM": "modeling_aurora.AuroraForCausalLM"
},
"model_type": "aurora",
"model_name": "Aurora Proelia ChatML",
"vocab_size": 16000,
"hidden_size": 896,
"num_layers": 23,
"num_attention_heads": 14,
"num_key_value_heads": 2,
"intermediate_size": 2432,
"context_length": 2048,
"max_position_embeddings": 2048,
"rope_theta": 500000.0,
"rms_norm_eps": 0.00001,
"qk_norm": true,
"tie_word_embeddings": true,
"attention_bias": false,
"mlp_bias": false,
"dropout": 0.0,
"num_experts": 1,
"router_aux_loss_coef": 0.0,
"router_z_loss_coef": 0.0,
"router_noise_scale": 0.0,
"moe_capacity_factor": 0.0,
"router_use_gate_weight": false,
"torch_dtype": "float16",
"library_name": "aurora"
}