Toastmachine/exaone_CSAT_test
09
1{2 "activation_function": "silu",3 "architectures": [4 "ExaoneForCausalLM"5 ],6 "attention_dropout": 0.0,7 "auto_map": {8 "AutoConfig": "configuration_exaone.ExaoneConfig",9 "AutoModelForCausalLM": "modeling_exaone.ExaoneForCausalLM",10 "AutoModelForSequenceClassification": "modeling_exaone.ExaoneForSequenceClassification"11 },12 "bos_token_id": 1,13 "embed_dropout": 0.0,14 "eos_token_id": 361,15 "hidden_size": 4096,16 "initializer_range": 0.02,17 "intermediate_size": 14336,18 "layer_norm_epsilon": 1e-05,19 "max_position_embeddings": 4096,20 "model_type": "exaone",21 "num_attention_heads": 32,22 "num_key_value_heads": 8,23 "num_layers": 32,24 "pad_token_id": 0,25 "rope_scaling": null,26 "rope_theta": 500000.0,27 "tie_word_embeddings": false,28 "torch_dtype": "float32",29 "transformers_version": "4.41.0",30 "use_cache": true,31 "vocab_size": 10240032}33 