config.json

{
    "architectures": [
        "Exaone4ForCausalLM"
    ],
    "attention_dropout": 0.0,
    "bos_token_id": 1,
    "torch_dtype": "float16",
    "eos_token_id": 361,
    "head_dim": 64,
    "hidden_act": "silu",
    "hidden_size": 2048,
    "initializer_range": 0.02,
    "intermediate_size": 4096,
    "layer_types": [
        "full_attention",
        "full_attention",
        "full_attention",
        "full_attention",
        "full_attention",
        "full_attention",
        "full_attention",
        "full_attention",
        "full_attention",
        "full_attention",
        "full_attention",
        "full_attention",
        "full_attention",
        "full_attention",
        "full_attention",
        "full_attention",
        "full_attention",
        "full_attention",
        "full_attention",
        "full_attention",
        "full_attention",
        "full_attention",
        "full_attention",
        "full_attention",
        "full_attention",
        "full_attention",
        "full_attention",
        "full_attention",
        "full_attention",
        "full_attention"
    ],
    "max_position_embeddings": 65536,
    "model_type": "exaone4",
    "num_attention_heads": 32,
    "num_hidden_layers": 30,
    "num_key_value_heads": 8,
    "pad_token_id": 0,
    "rms_norm_eps": 1e-05,
    "rope_scaling": {
        "factor": 16.0,
        "high_freq_factor": 4.0,
        "low_freq_factor": 1.0,
        "original_max_position_embeddings": 8192,
        "rope_type": "llama3"
    },
    "rope_theta": 1000000.0,
    "sliding_window": null,
    "sliding_window_pattern": null,
    "tie_word_embeddings": true,
    "transformers_version": "4.56.2",
    "unsloth_version": "2025.10.9",
    "use_cache": true,
    "vocab_size": 102400
}