{ "architectures": [ "Gemma3ForConditionalGeneration" ], "boi_token_index": 255999, "eoi_token_index": 256000, "image_token_index": 262144, "initializer_range": 0.02, "is_vision_model": true, "mm_tokens_per_image": 256, "model_type": "gemma3", "text_config": { "attention_bias": false, "attention_dropout": 0.0, "attn_logit_softcapping": null, "bos_token_id": 1, "cache_implementation": "hybrid", "eos_token_id": 2, "final_logit_softcapping": null, "head_dim": 256, "hidden_act": "gelu", "hidden_activation": "gelu_pytorch_tanh", "hidden_size": 8, "initializer_range": 0.02, "intermediate_size": 8, "max_position_embeddings": 2048, "model_type": "gemma3_text", "num_attention_heads": 2, "num_hidden_layers": 1, "num_key_value_heads": 2, "query_pre_attn_scalar": 256, "rms_norm_eps": 1e-06, "rope_local_base_freq": 10000.0, "rope_scaling": null, "rope_theta": 10000.0, "sliding_window": 4096, "sliding_window_pattern": 6, "tie_word_embeddings": false, "use_cache": true, "vocab_size": 262144 }, "tie_word_embeddings": false, "torch_dtype": "bfloat16", "transformers_version": "4.51.3", "vision_config": { "attention_dropout": 0.0, "dropout": 0.0, "hidden_act": "quick_gelu", "hidden_size": 8, "image_size": 896, "initializer_factor": 1.0, "initializer_range": 0.02, "intermediate_size": 8, "layer_norm_eps": 1e-06, "model_type": "siglip_vision_model", "num_attention_heads": 2, "num_channels": 3, "num_hidden_layers": 2, "patch_size": 28, "use_absolute_position_embeddings": true, "use_relative_position_embeddings": false } }