llama-265m / config.json
JuncaiL's picture
fix state_dict loading in MoE model
3240d88 verified
{
"_name_or_path": "JuncaiL/llama-265m",
"add_weight_norm": false,
"architectures": [
"LlamaMoEForCausalLM"
],
"auto_map": {
"AutoConfig": "configuration_llama_moe.LlamaMoEConfig",
"AutoModel": "modeling_llama_moe_hf.LlamaMoEModel",
"AutoModelForCausalLM": "modeling_llama_moe_hf.LlamaMoEForCausalLM"
},
"bos_token_id": 0,
"calculator_type": "UniversalCalculator",
"capacity_factor": 1.25,
"drop_tokens": true,
"dropped_padding": "zero",
"eos_token_id": 2,
"gate_add_noise": true,
"gate_balance_loss_weight": 0.01,
"gate_network": "mlp",
"gate_noise_epsilon": 0.01,
"gate_type": "TopKBalancedNoisyGate",
"gate_use_balance": true,
"gate_use_softmax": true,
"gates": "mlp",
"hidden_act": "silu",
"hidden_size": 2048,
"initializer_range": 0.02,
"intermediate_size": 8192,
"max_position_embeddings": 2048,
"model_type": "llama_moe",
"multiply_gate_scores": true,
"num_attention_heads": 16,
"num_experts": 1,
"num_hidden_layers": 2,
"num_key_value_heads": 16,
"num_selects": 1,
"pad_token_id": 1,
"pretraining_tp": 1,
"rms_norm_eps": 1e-05,
"rope_scaling": null,
"score_scale_factor": 4.0,
"size_experts": [
[
8192
],
[
8192
]
],
"tie_word_embeddings": false,
"torch_dtype": "float32",
"transformers_version": "4.30.2",
"use_cache": true,
"vocab_size": 32000
}