config.json

{
    "architectures": [
        "Gemma3nForConditionalGeneration"
    ],
    "audio_config": {
        "conf_attention_chunk_size": 12,
        "conf_attention_context_left": 13,
        "conf_attention_context_right": 0,
        "conf_attention_logit_cap": 50.0,
        "conf_conv_kernel_size": 5,
        "conf_num_attention_heads": 8,
        "conf_num_hidden_layers": 12,
        "conf_positional_bias_size": 256,
        "conf_reduction_factor": 4,
        "conf_residual_weight": 0.5,
        "gradient_clipping": 10000000000.0,
        "hidden_size": 1536,
        "input_feat_size": 128,
        "model_type": "gemma3n_audio",
        "rms_norm_eps": 1e-06,
        "sscp_conv_channel_size": [
            128,
            32
        ],
        "sscp_conv_eps": 0.001,
        "sscp_conv_kernel_size": [
            [
                3,
                3
            ],
            [
                3,
                3
            ]
        ],
        "sscp_conv_stride_size": [
            [
                2,
                2
            ],
            [
                2,
                2
            ]
        ],
        "torch_dtype": "bfloat16",
        "vocab_offset": 262272,
        "vocab_size": 128
    },
    "audio_soft_tokens_per_image": 188,
    "audio_token_id": 262273,
    "boa_token_id": 256000,
    "boi_token_id": 255999,
    "eoa_token_id": 262272,
    "eoi_token_id": 262144,
    "eos_token_id": [
        1,
        106
    ],
    "image_token_id": 262145,
    "initializer_range": 0.02,
    "model_type": "gemma3n",
    "text_config": {
        "activation_sparsity_pattern": [
            0.95,
            0.95,
            0.95,
            0.95,
            0.95,
            0.95,
            0.95,
            0.95,
            0.95,
            0.95,
            0.0,
            0.0,
            0.0,
            0.0,
            0.0,
            0.0,
            0.0,
            0.0,
            0.0,
            0.0,
            0.0,
            0.0,
            0.0,
            0.0,
            0.0,
            0.0,
            0.0,
            0.0,
            0.0,
            0.0,
            0.0,
            0.0,
            0.0,
            0.0,
            0.0
        ],
        "altup_active_idx": 0,
        "altup_coef_clip": 120.0,
        "altup_correct_scale": true,
        "altup_lr_multiplier": 1.0,
        "altup_num_inputs": 4,
        "attention_bias": false,
        "attention_dropout": 0.0,
        "final_logit_softcapping": 30.0,
        "head_dim": 256,
        "hidden_activation": "gelu_pytorch_tanh",
        "hidden_size": 2048,
        "hidden_size_per_layer_input": 256,
        "initializer_range": 0.02,
        "intermediate_size": 16384,
        "laurel_rank": 64,
        "layer_types": [
            "sliding_attention",
            "sliding_attention",
            "sliding_attention",
            "sliding_attention",
            "full_attention",
            "sliding_attention",
            "sliding_attention",
            "sliding_attention",
            "sliding_attention",
            "full_attention",
            "sliding_attention",
            "sliding_attention",
            "sliding_attention",
            "sliding_attention",
            "full_attention",
            "sliding_attention",
            "sliding_attention",
            "sliding_attention",
            "sliding_attention",
            "full_attention",
            "sliding_attention",
            "sliding_attention",
            "sliding_attention",
            "sliding_attention",
            "full_attention",
            "sliding_attention",
            "sliding_attention",
            "sliding_attention",
            "sliding_attention",
            "full_attention",
            "sliding_attention",
            "sliding_attention",
            "sliding_attention",
            "sliding_attention",
            "full_attention"
        ],
        "max_position_embeddings": 32768,
        "model_type": "gemma3n_text",
        "num_attention_heads": 8,
        "num_hidden_layers": 35,
        "num_key_value_heads": 2,
        "num_kv_shared_layers": 15,
        "query_pre_attn_scalar": 256,
        "rms_norm_eps": 1e-06,
        "rope_local_base_freq": 10000.0,
        "rope_scaling": null,
        "rope_theta": 1000000.0,
        "sliding_window": 512,
        "torch_dtype": "bfloat16",
        "use_cache": true,
        "vocab_size": 262400,
        "vocab_size_per_layer_input": 262144
    },
    "transformers_version": "4.53.0.dev0",
    "vision_config": {
        "architecture": "mobilenetv5_300m_enc",
        "do_pooling": true,
        "hidden_size": 2048,
        "initializer_range": 0.02,
        "label_names": [
            "LABEL_0",
            "LABEL_1"
        ],
        "model_type": "gemma3n_vision",
        "num_classes": 2,
        "rms_norm_eps": 1e-06,
        "torch_dtype": "bfloat16",
        "vocab_offset": 262144,
        "vocab_size": 128
    },
    "vision_soft_tokens_per_image": 256
}