DarkSca/GenZ_BitNet
018
1{2 "architectures": [3 "BitNetForCausalLM"4 ],5 "attention_bias": false,6 "attention_dropout": 0.0,7 "auto_map": {8 "AutoConfig": "configuration_bitnet.BitNetConfig",9 "AutoModelForCausalLM": "modeling_bitnet.BitNetForCausalLM"10 },11 "bos_token_id": 128000,12 "dtype": "float32",13 "eos_token_id": 128001,14 "hidden_act": "relu2",15 "hidden_size": 2560,16 "initializer_range": 0.02,17 "intermediate_size": 6912,18 "max_position_embeddings": 4096,19 "model_type": "bitnet",20 "num_attention_heads": 20,21 "num_hidden_layers": 30,22 "num_key_value_heads": 5,23 "quantization_config": {24 "linear_class": "autobitlinear",25 "modules_to_not_convert": null,26 "quant_method": "bitnet",27 "quantization_mode": "online",28 "rms_norm_eps": 1e-06,29 "use_rms_norm": false30 },31 "rms_norm_eps": 1e-05,32 "rope_theta": 500000.0,33 "tie_word_embeddings": true,34 "transformers_version": "4.57.1",35 "use_cache": true,36 "vocab_size": 12825637}38 