{ "architectures": [ "DiffusionGemmaForBlockDiffusion" ], "boi_token_id": 255999, "canvas_length": 256, "dtype": "bfloat16", "eoi_token_id": 258882, "eos_token_id": [ 1, 106 ], "has_audio": false, "has_video": true, "has_vision": true, "image_token_id": 258880, "initializer_range": 0.02, "modalities": { "audio": false, "image": true, "text": true, "video": true, "vision": true }, "modality_notes": { "audio": "processor_config has feature_extractor metadata, but config has no audio_config/audio_token_id", "image": "vision_config plus image_token_id are present; runtime proof still required", "video": "processor_config has video_processor, but config has no video_token_id; proof required" }, "model_type": "diffusion_gemma", "quantization": { "bits": 8, "canvas_length": 256, "family": "diffusion_gemma", "group_size": 32, "mode": "mxfp8", "model.decoder.embed_tokens": false, "model.decoder.layers.0.input_layernorm": false, "model.decoder.layers.0.post_attention_layernorm": false, "model.decoder.layers.0.post_feedforward_layernorm": false, "model.decoder.layers.0.post_feedforward_layernorm_1": false, "model.decoder.layers.0.post_feedforward_layernorm_2": false, "model.decoder.layers.0.pre_feedforward_layernorm": false, "model.decoder.layers.0.pre_feedforward_layernorm_2": false, "model.decoder.layers.0.self_attn.k_norm": false, "model.decoder.layers.0.self_attn.q_norm": false, "model.decoder.layers.1.input_layernorm": false, "model.decoder.layers.1.post_attention_layernorm": false, "model.decoder.layers.1.post_feedforward_layernorm": false, "model.decoder.layers.1.post_feedforward_layernorm_1": false, "model.decoder.layers.1.post_feedforward_layernorm_2": false, "model.decoder.layers.1.pre_feedforward_layernorm": false, "model.decoder.layers.1.pre_feedforward_layernorm_2": false, "model.decoder.layers.1.self_attn.k_norm": false, "model.decoder.layers.1.self_attn.q_norm": false, "model.decoder.layers.10.input_layernorm": false, "model.decoder.layers.10.post_attention_layernorm": false, "model.decoder.layers.10.post_feedforward_layernorm": false, "model.decoder.layers.10.post_feedforward_layernorm_1": false, "model.decoder.layers.10.post_feedforward_layernorm_2": false, "model.decoder.layers.10.pre_feedforward_layernorm": false, "model.decoder.layers.10.pre_feedforward_layernorm_2": false, "model.decoder.layers.10.self_attn.k_norm": false, "model.decoder.layers.10.self_attn.q_norm": false, "model.decoder.layers.11.input_layernorm": false, "model.decoder.layers.11.post_attention_layernorm": false, "model.decoder.layers.11.post_feedforward_layernorm": false, "model.decoder.layers.11.post_feedforward_layernorm_1": false, "model.decoder.layers.11.post_feedforward_layernorm_2": false, "model.decoder.layers.11.pre_feedforward_layernorm": false, "model.decoder.layers.11.pre_feedforward_layernorm_2": false, "model.decoder.layers.11.self_attn.k_norm": false, "model.decoder.layers.11.self_attn.q_norm": false, "model.decoder.layers.12.input_layernorm": false, "model.decoder.layers.12.post_attention_layernorm": false, "model.decoder.layers.12.post_feedforward_layernorm": false, "model.decoder.layers.12.post_feedforward_layernorm_1": false, "model.decoder.layers.12.post_feedforward_layernorm_2": false, "model.decoder.layers.12.pre_feedforward_layernorm": false, "model.decoder.layers.12.pre_feedforward_layernorm_2": false, "model.decoder.layers.12.self_attn.k_norm": false, "model.decoder.layers.12.self_attn.q_norm": false, "model.decoder.layers.13.input_layernorm": false, "model.decoder.layers.13.post_attention_layernorm": false, "model.decoder.layers.13.post_feedforward_layernorm": false, "model.decoder.layers.13.post_feedforward_layernorm_1": false, "model.decoder.layers.13.post_feedforward_layernorm_2": false, "model.decoder.layers.13.pre_feedforward_layernorm": false, "model.decoder.layers.13.pre_feedforward_layernorm_2": false, "model.decoder.layers.13.self_attn.k_norm": false, "model.decoder.layers.13.self_attn.q_norm": false, "model.decoder.layers.14.input_layernorm": false, "model.decoder.layers.14.post_attention_layernorm": false, "model.decoder.layers.14.post_feedforward_layernorm": false, "model.decoder.layers.14.post_feedforward_layernorm_1": false, "model.decoder.layers.14.post_feedforward_layernorm_2": false, "model.decoder.layers.14.pre_feedforward_layernorm": false, "model.decoder.layers.14.pre_feedforward_layernorm_2": false, "model.decoder.layers.14.self_attn.k_norm": false, "model.decoder.layers.14.self_attn.q_norm": false, "model.decoder.layers.15.input_layernorm": false, "model.decoder.layers.15.post_attention_layernorm": false, "model.decoder.layers.15.post_feedforward_layernorm": false, "model.decoder.layers.15.post_feedforward_layernorm_1": false, "model.decoder.layers.15.post_feedforward_layernorm_2": false, "model.decoder.layers.15.pre_feedforward_layernorm": false, "model.decoder.layers.15.pre_feedforward_layernorm_2": false, "model.decoder.layers.15.self_attn.k_norm": false, "model.decoder.layers.15.self_attn.q_norm": false, "model.decoder.layers.16.input_layernorm": false, "model.decoder.layers.16.post_attention_layernorm": false, "model.decoder.layers.16.post_feedforward_layernorm": false, "model.decoder.layers.16.post_feedforward_layernorm_1": false, "model.decoder.layers.16.post_feedforward_layernorm_2": false, "model.decoder.layers.16.pre_feedforward_layernorm": false, "model.decoder.layers.16.pre_feedforward_layernorm_2": false, "model.decoder.layers.16.self_attn.k_norm": false, "model.decoder.layers.16.self_attn.q_norm": false, "model.decoder.layers.17.input_layernorm": false, "model.decoder.layers.17.post_attention_layernorm": false, "model.decoder.layers.17.post_feedforward_layernorm": false, "model.decoder.layers.17.post_feedforward_layernorm_1": false, "model.decoder.layers.17.post_feedforward_layernorm_2": false, "model.decoder.layers.17.pre_feedforward_layernorm": false, "model.decoder.layers.17.pre_feedforward_layernorm_2": false, "model.decoder.layers.17.self_attn.k_norm": false, "model.decoder.layers.17.self_attn.q_norm": false, "model.decoder.layers.18.input_layernorm": false, "model.decoder.layers.18.post_attention_layernorm": false, "model.decoder.layers.18.post_feedforward_layernorm": false, "model.decoder.layers.18.post_feedforward_layernorm_1": false, "model.decoder.layers.18.post_feedforward_layernorm_2": false, "model.decoder.layers.18.pre_feedforward_layernorm": false, "model.decoder.layers.18.pre_feedforward_layernorm_2": false, "model.decoder.layers.18.self_attn.k_norm": false, "model.decoder.layers.18.self_attn.q_norm": false, "model.decoder.layers.19.input_layernorm": false, "model.decoder.layers.19.post_attention_layernorm": false, "model.decoder.layers.19.post_feedforward_layernorm": false, "model.decoder.layers.19.post_feedforward_layernorm_1": false, "model.decoder.layers.19.post_feedforward_layernorm_2": false, "model.decoder.layers.19.pre_feedforward_layernorm": false, "model.decoder.layers.19.pre_feedforward_layernorm_2": false, "model.decoder.layers.19.self_attn.k_norm": false, "model.decoder.layers.19.self_attn.q_norm": false, "model.decoder.layers.2.input_layernorm": false, "model.decoder.layers.2.post_attention_layernorm": false, "model.decoder.layers.2.post_feedforward_layernorm": false, "model.decoder.layers.2.post_feedforward_layernorm_1": false, "model.decoder.layers.2.post_feedforward_layernorm_2": false, "model.decoder.layers.2.pre_feedforward_layernorm": false, "model.decoder.layers.2.pre_feedforward_layernorm_2": false, "model.decoder.layers.2.self_attn.k_norm": false, "model.decoder.layers.2.self_attn.q_norm": false, "model.decoder.layers.20.input_layernorm": false, "model.decoder.layers.20.post_attention_layernorm": false, "model.decoder.layers.20.post_feedforward_layernorm": false, "model.decoder.layers.20.post_feedforward_layernorm_1": false, "model.decoder.layers.20.post_feedforward_layernorm_2": false, "model.decoder.layers.20.pre_feedforward_layernorm": false, "model.decoder.layers.20.pre_feedforward_layernorm_2": false, "model.decoder.layers.20.self_attn.k_norm": false, "model.decoder.layers.20.self_attn.q_norm": false, "model.decoder.layers.21.input_layernorm": false, "model.decoder.layers.21.post_attention_layernorm": false, "model.decoder.layers.21.post_feedforward_layernorm": false, "model.decoder.layers.21.post_feedforward_layernorm_1": false, "model.decoder.layers.21.post_feedforward_layernorm_2": false, "model.decoder.layers.21.pre_feedforward_layernorm": false, "model.decoder.layers.21.pre_feedforward_layernorm_2": false, "model.decoder.layers.21.self_attn.k_norm": false, "model.decoder.layers.21.self_attn.q_norm": false, "model.decoder.layers.22.input_layernorm": false, "model.decoder.layers.22.post_attention_layernorm": false, "model.decoder.layers.22.post_feedforward_layernorm": false, "model.decoder.layers.22.post_feedforward_layernorm_1": false, "model.decoder.layers.22.post_feedforward_layernorm_2": false, "model.decoder.layers.22.pre_feedforward_layernorm": false, "model.decoder.layers.22.pre_feedforward_layernorm_2": false, "model.decoder.layers.22.self_attn.k_norm": false, "model.decoder.layers.22.self_attn.q_norm": false, "model.decoder.layers.23.input_layernorm": false, "model.decoder.layers.23.post_attention_layernorm": false, "model.decoder.layers.23.post_feedforward_layernorm": false, "model.decoder.layers.23.post_feedforward_layernorm_1": false, "model.decoder.layers.23.post_feedforward_layernorm_2": false, "model.decoder.layers.23.pre_feedforward_layernorm": false, "model.decoder.layers.23.pre_feedforward_layernorm_2": false, "model.decoder.layers.23.self_attn.k_norm": false, "model.decoder.layers.23.self_attn.q_norm": false, "model.decoder.layers.24.input_layernorm": false, "model.decoder.layers.24.post_attention_layernorm": false, "model.decoder.layers.24.post_feedforward_layernorm": false, "model.decoder.layers.24.post_feedforward_layernorm_1": false, "model.decoder.layers.24.post_feedforward_layernorm_2": false, "model.decoder.layers.24.pre_feedforward_layernorm": false, "model.decoder.layers.24.pre_feedforward_layernorm_2": false, "model.decoder.layers.24.self_attn.k_norm": false, "model.decoder.layers.24.self_attn.q_norm": false, "model.decoder.layers.25.input_layernorm": false, "model.decoder.layers.25.post_attention_layernorm": false, "model.decoder.layers.25.post_feedforward_layernorm": false, "model.decoder.layers.25.post_feedforward_layernorm_1": false, "model.decoder.layers.25.post_feedforward_layernorm_2": false, "model.decoder.layers.25.pre_feedforward_layernorm": false, "model.decoder.layers.25.pre_feedforward_layernorm_2": false, "model.decoder.layers.25.self_attn.k_norm": false, "model.decoder.layers.25.self_attn.q_norm": false, "model.decoder.layers.26.input_layernorm": false, "model.decoder.layers.26.post_attention_layernorm": false, "model.decoder.layers.26.post_feedforward_layernorm": false, "model.decoder.layers.26.post_feedforward_layernorm_1": false, "model.decoder.layers.26.post_feedforward_layernorm_2": false, "model.decoder.layers.26.pre_feedforward_layernorm": false, "model.decoder.layers.26.pre_feedforward_layernorm_2": false, "model.decoder.layers.26.self_attn.k_norm": false, "model.decoder.layers.26.self_attn.q_norm": false, "model.decoder.layers.27.input_layernorm": false, "model.decoder.layers.27.post_attention_layernorm": false, "model.decoder.layers.27.post_feedforward_layernorm": false, "model.decoder.layers.27.post_feedforward_layernorm_1": false, "model.decoder.layers.27.post_feedforward_layernorm_2": false, "model.decoder.layers.27.pre_feedforward_layernorm": false, "model.decoder.layers.27.pre_feedforward_layernorm_2": false, "model.decoder.layers.27.self_attn.k_norm": false, "model.decoder.layers.27.self_attn.q_norm": false, "model.decoder.layers.28.input_layernorm": false, "model.decoder.layers.28.post_attention_layernorm": false, "model.decoder.layers.28.post_feedforward_layernorm": false, "model.decoder.layers.28.post_feedforward_layernorm_1": false, "model.decoder.layers.28.post_feedforward_layernorm_2": false, "model.decoder.layers.28.pre_feedforward_layernorm": false, "model.decoder.layers.28.pre_feedforward_layernorm_2": false, "model.decoder.layers.28.self_attn.k_norm": false, "model.decoder.layers.28.self_attn.q_norm": false, "model.decoder.layers.29.input_layernorm": false, "model.decoder.layers.29.post_attention_layernorm": false, "model.decoder.layers.29.post_feedforward_layernorm": false, "model.decoder.layers.29.post_feedforward_layernorm_1": false, "model.decoder.layers.29.post_feedforward_layernorm_2": false, "model.decoder.layers.29.pre_feedforward_layernorm": false, "model.decoder.layers.29.pre_feedforward_layernorm_2": false, "model.decoder.layers.29.self_attn.k_norm": false, "model.decoder.layers.29.self_attn.q_norm": false, "model.decoder.layers.3.input_layernorm": false, "model.decoder.layers.3.post_attention_layernorm": false, "model.decoder.layers.3.post_feedforward_layernorm": false, "model.decoder.layers.3.post_feedforward_layernorm_1": false, "model.decoder.layers.3.post_feedforward_layernorm_2": false, "model.decoder.layers.3.pre_feedforward_layernorm": false, "model.decoder.layers.3.pre_feedforward_layernorm_2": false, "model.decoder.layers.3.self_attn.k_norm": false, "model.decoder.layers.3.self_attn.q_norm": false, "model.decoder.layers.4.input_layernorm": false, "model.decoder.layers.4.post_attention_layernorm": false, "model.decoder.layers.4.post_feedforward_layernorm": false, "model.decoder.layers.4.post_feedforward_layernorm_1": false, "model.decoder.layers.4.post_feedforward_layernorm_2": false, "model.decoder.layers.4.pre_feedforward_layernorm": false, "model.decoder.layers.4.pre_feedforward_layernorm_2": false, "model.decoder.layers.4.self_attn.k_norm": false, "model.decoder.layers.4.self_attn.q_norm": false, "model.decoder.layers.5.input_layernorm": false, "model.decoder.layers.5.post_attention_layernorm": false, "model.decoder.layers.5.post_feedforward_layernorm": false, "model.decoder.layers.5.post_feedforward_layernorm_1": false, "model.decoder.layers.5.post_feedforward_layernorm_2": false, "model.decoder.layers.5.pre_feedforward_layernorm": false, "model.decoder.layers.5.pre_feedforward_layernorm_2": false, "model.decoder.layers.5.self_attn.k_norm": false, "model.decoder.layers.5.self_attn.q_norm": false, "model.decoder.layers.6.input_layernorm": false, "model.decoder.layers.6.post_attention_layernorm": false, "model.decoder.layers.6.post_feedforward_layernorm": false, "model.decoder.layers.6.post_feedforward_layernorm_1": false, "model.decoder.layers.6.post_feedforward_layernorm_2": false, "model.decoder.layers.6.pre_feedforward_layernorm": false, "model.decoder.layers.6.pre_feedforward_layernorm_2": false, "model.decoder.layers.6.self_attn.k_norm": false, "model.decoder.layers.6.self_attn.q_norm": false, "model.decoder.layers.7.input_layernorm": false, "model.decoder.layers.7.post_attention_layernorm": false, "model.decoder.layers.7.post_feedforward_layernorm": false, "model.decoder.layers.7.post_feedforward_layernorm_1": false, "model.decoder.layers.7.post_feedforward_layernorm_2": false, "model.decoder.layers.7.pre_feedforward_layernorm": false, "model.decoder.layers.7.pre_feedforward_layernorm_2": false, "model.decoder.layers.7.self_attn.k_norm": false, "model.decoder.layers.7.self_attn.q_norm": false, "model.decoder.layers.8.input_layernorm": false, "model.decoder.layers.8.post_attention_layernorm": false, "model.decoder.layers.8.post_feedforward_layernorm": false, "model.decoder.layers.8.post_feedforward_layernorm_1": false, "model.decoder.layers.8.post_feedforward_layernorm_2": false, "model.decoder.layers.8.pre_feedforward_layernorm": false, "model.decoder.layers.8.pre_feedforward_layernorm_2": false, "model.decoder.layers.8.self_attn.k_norm": false, "model.decoder.layers.8.self_attn.q_norm": false, "model.decoder.layers.9.input_layernorm": false, "model.decoder.layers.9.post_attention_layernorm": false, "model.decoder.layers.9.post_feedforward_layernorm": false, "model.decoder.layers.9.post_feedforward_layernorm_1": false, "model.decoder.layers.9.post_feedforward_layernorm_2": false, "model.decoder.layers.9.pre_feedforward_layernorm": false, "model.decoder.layers.9.pre_feedforward_layernorm_2": false, "model.decoder.layers.9.self_attn.k_norm": false, "model.decoder.layers.9.self_attn.q_norm": false, "model.decoder.norm": false, "model.decoder.self_conditioning.down_proj": false, "model.decoder.self_conditioning.gate_proj": false, "model.decoder.self_conditioning.pre_norm": false, "model.decoder.self_conditioning.up_proj": false, "model.encoder.embed_vision.embedding_projection": false, "model.encoder.vision_tower.encoder.layers.0.input_layernorm": false, "model.encoder.vision_tower.encoder.layers.0.mlp.down_proj.linear": false, "model.encoder.vision_tower.encoder.layers.0.mlp.gate_proj.linear": false, "model.encoder.vision_tower.encoder.layers.0.mlp.up_proj.linear": false, "model.encoder.vision_tower.encoder.layers.0.post_attention_layernorm": false, "model.encoder.vision_tower.encoder.layers.0.post_feedforward_layernorm": false, "model.encoder.vision_tower.encoder.layers.0.pre_feedforward_layernorm": false, "model.encoder.vision_tower.encoder.layers.0.self_attn.k_norm": false, "model.encoder.vision_tower.encoder.layers.0.self_attn.k_proj.linear": false, "model.encoder.vision_tower.encoder.layers.0.self_attn.o_proj.linear": false, "model.encoder.vision_tower.encoder.layers.0.self_attn.q_norm": false, "model.encoder.vision_tower.encoder.layers.0.self_attn.q_proj.linear": false, "model.encoder.vision_tower.encoder.layers.0.self_attn.v_proj.linear": false, "model.encoder.vision_tower.encoder.layers.1.input_layernorm": false, "model.encoder.vision_tower.encoder.layers.1.mlp.down_proj.linear": false, "model.encoder.vision_tower.encoder.layers.1.mlp.gate_proj.linear": false, "model.encoder.vision_tower.encoder.layers.1.mlp.up_proj.linear": false, "model.encoder.vision_tower.encoder.layers.1.post_attention_layernorm": false, "model.encoder.vision_tower.encoder.layers.1.post_feedforward_layernorm": false, "model.encoder.vision_tower.encoder.layers.1.pre_feedforward_layernorm": false, "model.encoder.vision_tower.encoder.layers.1.self_attn.k_norm": false, "model.encoder.vision_tower.encoder.layers.1.self_attn.k_proj.linear": false, "model.encoder.vision_tower.encoder.layers.1.self_attn.o_proj.linear": false, "model.encoder.vision_tower.encoder.layers.1.self_attn.q_norm": false, "model.encoder.vision_tower.encoder.layers.1.self_attn.q_proj.linear": false, "model.encoder.vision_tower.encoder.layers.1.self_attn.v_proj.linear": false, "model.encoder.vision_tower.encoder.layers.10.input_layernorm": false, "model.encoder.vision_tower.encoder.layers.10.mlp.down_proj.linear": false, "model.encoder.vision_tower.encoder.layers.10.mlp.gate_proj.linear": false, "model.encoder.vision_tower.encoder.layers.10.mlp.up_proj.linear": false, "model.encoder.vision_tower.encoder.layers.10.post_attention_layernorm": false, "model.encoder.vision_tower.encoder.layers.10.post_feedforward_layernorm": false, "model.encoder.vision_tower.encoder.layers.10.pre_feedforward_layernorm": false, "model.encoder.vision_tower.encoder.layers.10.self_attn.k_norm": false, "model.encoder.vision_tower.encoder.layers.10.self_attn.k_proj.linear": false, "model.encoder.vision_tower.encoder.layers.10.self_attn.o_proj.linear": false, "model.encoder.vision_tower.encoder.layers.10.self_attn.q_norm": false, "model.encoder.vision_tower.encoder.layers.10.self_attn.q_proj.linear": false, "model.encoder.vision_tower.encoder.layers.10.self_attn.v_proj.linear": false, "model.encoder.vision_tower.encoder.layers.11.input_layernorm": false, "model.encoder.vision_tower.encoder.layers.11.mlp.down_proj.linear": false, "model.encoder.vision_tower.encoder.layers.11.mlp.gate_proj.linear": false, "model.encoder.vision_tower.encoder.layers.11.mlp.up_proj.linear": false, "model.encoder.vision_tower.encoder.layers.11.post_attention_layernorm": false, "model.encoder.vision_tower.encoder.layers.11.post_feedforward_layernorm": false, "model.encoder.vision_tower.encoder.layers.11.pre_feedforward_layernorm": false, "model.encoder.vision_tower.encoder.layers.11.self_attn.k_norm": false, "model.encoder.vision_tower.encoder.layers.11.self_attn.k_proj.linear": false, "model.encoder.vision_tower.encoder.layers.11.self_attn.o_proj.linear": false, "model.encoder.vision_tower.encoder.layers.11.self_attn.q_norm": false, "model.encoder.vision_tower.encoder.layers.11.self_attn.q_proj.linear": false, "model.encoder.vision_tower.encoder.layers.11.self_attn.v_proj.linear": false, "model.encoder.vision_tower.encoder.layers.12.input_layernorm": false, "model.encoder.vision_tower.encoder.layers.12.mlp.down_proj.linear": false, "model.encoder.vision_tower.encoder.layers.12.mlp.gate_proj.linear": false, "model.encoder.vision_tower.encoder.layers.12.mlp.up_proj.linear": false, "model.encoder.vision_tower.encoder.layers.12.post_attention_layernorm": false, "model.encoder.vision_tower.encoder.layers.12.post_feedforward_layernorm": false, "model.encoder.vision_tower.encoder.layers.12.pre_feedforward_layernorm": false, "model.encoder.vision_tower.encoder.layers.12.self_attn.k_norm": false, "model.encoder.vision_tower.encoder.layers.12.self_attn.k_proj.linear": false, "model.encoder.vision_tower.encoder.layers.12.self_attn.o_proj.linear": false, "model.encoder.vision_tower.encoder.layers.12.self_attn.q_norm": false, "model.encoder.vision_tower.encoder.layers.12.self_attn.q_proj.linear": false, "model.encoder.vision_tower.encoder.layers.12.self_attn.v_proj.linear": false, "model.encoder.vision_tower.encoder.layers.13.input_layernorm": false, "model.encoder.vision_tower.encoder.layers.13.mlp.down_proj.linear": false, "model.encoder.vision_tower.encoder.layers.13.mlp.gate_proj.linear": false, "model.encoder.vision_tower.encoder.layers.13.mlp.up_proj.linear": false, "model.encoder.vision_tower.encoder.layers.13.post_attention_layernorm": false, "model.encoder.vision_tower.encoder.layers.13.post_feedforward_layernorm": false, "model.encoder.vision_tower.encoder.layers.13.pre_feedforward_layernorm": false, "model.encoder.vision_tower.encoder.layers.13.self_attn.k_norm": false, "model.encoder.vision_tower.encoder.layers.13.self_attn.k_proj.linear": false, "model.encoder.vision_tower.encoder.layers.13.self_attn.o_proj.linear": false, "model.encoder.vision_tower.encoder.layers.13.self_attn.q_norm": false, "model.encoder.vision_tower.encoder.layers.13.self_attn.q_proj.linear": false, "model.encoder.vision_tower.encoder.layers.13.self_attn.v_proj.linear": false, "model.encoder.vision_tower.encoder.layers.14.input_layernorm": false, "model.encoder.vision_tower.encoder.layers.14.mlp.down_proj.linear": false, "model.encoder.vision_tower.encoder.layers.14.mlp.gate_proj.linear": false, "model.encoder.vision_tower.encoder.layers.14.mlp.up_proj.linear": false, "model.encoder.vision_tower.encoder.layers.14.post_attention_layernorm": false, "model.encoder.vision_tower.encoder.layers.14.post_feedforward_layernorm": false, "model.encoder.vision_tower.encoder.layers.14.pre_feedforward_layernorm": false, "model.encoder.vision_tower.encoder.layers.14.self_attn.k_norm": false, "model.encoder.vision_tower.encoder.layers.14.self_attn.k_proj.linear": false, "model.encoder.vision_tower.encoder.layers.14.self_attn.o_proj.linear": false, "model.encoder.vision_tower.encoder.layers.14.self_attn.q_norm": false, "model.encoder.vision_tower.encoder.layers.14.self_attn.q_proj.linear": false, "model.encoder.vision_tower.encoder.layers.14.self_attn.v_proj.linear": false, "model.encoder.vision_tower.encoder.layers.15.input_layernorm": false, "model.encoder.vision_tower.encoder.layers.15.mlp.down_proj.linear": false, "model.encoder.vision_tower.encoder.layers.15.mlp.gate_proj.linear": false, "model.encoder.vision_tower.encoder.layers.15.mlp.up_proj.linear": false, "model.encoder.vision_tower.encoder.layers.15.post_attention_layernorm": false, "model.encoder.vision_tower.encoder.layers.15.post_feedforward_layernorm": false, "model.encoder.vision_tower.encoder.layers.15.pre_feedforward_layernorm": false, "model.encoder.vision_tower.encoder.layers.15.self_attn.k_norm": false, "model.encoder.vision_tower.encoder.layers.15.self_attn.k_proj.linear": false, "model.encoder.vision_tower.encoder.layers.15.self_attn.o_proj.linear": false, "model.encoder.vision_tower.encoder.layers.15.self_attn.q_norm": false, "model.encoder.vision_tower.encoder.layers.15.self_attn.q_proj.linear": false, "model.encoder.vision_tower.encoder.layers.15.self_attn.v_proj.linear": false, "model.encoder.vision_tower.encoder.layers.16.input_layernorm": false, "model.encoder.vision_tower.encoder.layers.16.mlp.down_proj.linear": false, "model.encoder.vision_tower.encoder.layers.16.mlp.gate_proj.linear": false, "model.encoder.vision_tower.encoder.layers.16.mlp.up_proj.linear": false, "model.encoder.vision_tower.encoder.layers.16.post_attention_layernorm": false, "model.encoder.vision_tower.encoder.layers.16.post_feedforward_layernorm": false, "model.encoder.vision_tower.encoder.layers.16.pre_feedforward_layernorm": false, "model.encoder.vision_tower.encoder.layers.16.self_attn.k_norm": false, "model.encoder.vision_tower.encoder.layers.16.self_attn.k_proj.linear": false, "model.encoder.vision_tower.encoder.layers.16.self_attn.o_proj.linear": false, "model.encoder.vision_tower.encoder.layers.16.self_attn.q_norm": false, "model.encoder.vision_tower.encoder.layers.16.self_attn.q_proj.linear": false, "model.encoder.vision_tower.encoder.layers.16.self_attn.v_proj.linear": false, "model.encoder.vision_tower.encoder.layers.17.input_layernorm": false, "model.encoder.vision_tower.encoder.layers.17.mlp.down_proj.linear": false, "model.encoder.vision_tower.encoder.layers.17.mlp.gate_proj.linear": false, "model.encoder.vision_tower.encoder.layers.17.mlp.up_proj.linear": false, "model.encoder.vision_tower.encoder.layers.17.post_attention_layernorm": false, "model.encoder.vision_tower.encoder.layers.17.post_feedforward_layernorm": false, "model.encoder.vision_tower.encoder.layers.17.pre_feedforward_layernorm": false, "model.encoder.vision_tower.encoder.layers.17.self_attn.k_norm": false, "model.encoder.vision_tower.encoder.layers.17.self_attn.k_proj.linear": false, "model.encoder.vision_tower.encoder.layers.17.self_attn.o_proj.linear": false, "model.encoder.vision_tower.encoder.layers.17.self_attn.q_norm": false, "model.encoder.vision_tower.encoder.layers.17.self_attn.q_proj.linear": false, "model.encoder.vision_tower.encoder.layers.17.self_attn.v_proj.linear": false, "model.encoder.vision_tower.encoder.layers.18.input_layernorm": false, "model.encoder.vision_tower.encoder.layers.18.mlp.down_proj.linear": false, "model.encoder.vision_tower.encoder.layers.18.mlp.gate_proj.linear": false, "model.encoder.vision_tower.encoder.layers.18.mlp.up_proj.linear": false, "model.encoder.vision_tower.encoder.layers.18.post_attention_layernorm": false, "model.encoder.vision_tower.encoder.layers.18.post_feedforward_layernorm": false, "model.encoder.vision_tower.encoder.layers.18.pre_feedforward_layernorm": false, "model.encoder.vision_tower.encoder.layers.18.self_attn.k_norm": false, "model.encoder.vision_tower.encoder.layers.18.self_attn.k_proj.linear": false, "model.encoder.vision_tower.encoder.layers.18.self_attn.o_proj.linear": false, "model.encoder.vision_tower.encoder.layers.18.self_attn.q_norm": false, "model.encoder.vision_tower.encoder.layers.18.self_attn.q_proj.linear": false, "model.encoder.vision_tower.encoder.layers.18.self_attn.v_proj.linear": false, "model.encoder.vision_tower.encoder.layers.19.input_layernorm": false, "model.encoder.vision_tower.encoder.layers.19.mlp.down_proj.linear": false, "model.encoder.vision_tower.encoder.layers.19.mlp.gate_proj.linear": false, "model.encoder.vision_tower.encoder.layers.19.mlp.up_proj.linear": false, "model.encoder.vision_tower.encoder.layers.19.post_attention_layernorm": false, "model.encoder.vision_tower.encoder.layers.19.post_feedforward_layernorm": false, "model.encoder.vision_tower.encoder.layers.19.pre_feedforward_layernorm": false, "model.encoder.vision_tower.encoder.layers.19.self_attn.k_norm": false, "model.encoder.vision_tower.encoder.layers.19.self_attn.k_proj.linear": false, "model.encoder.vision_tower.encoder.layers.19.self_attn.o_proj.linear": false, "model.encoder.vision_tower.encoder.layers.19.self_attn.q_norm": false, "model.encoder.vision_tower.encoder.layers.19.self_attn.q_proj.linear": false, "model.encoder.vision_tower.encoder.layers.19.self_attn.v_proj.linear": false, "model.encoder.vision_tower.encoder.layers.2.input_layernorm": false, "model.encoder.vision_tower.encoder.layers.2.mlp.down_proj.linear": false, "model.encoder.vision_tower.encoder.layers.2.mlp.gate_proj.linear": false, "model.encoder.vision_tower.encoder.layers.2.mlp.up_proj.linear": false, "model.encoder.vision_tower.encoder.layers.2.post_attention_layernorm": false, "model.encoder.vision_tower.encoder.layers.2.post_feedforward_layernorm": false, "model.encoder.vision_tower.encoder.layers.2.pre_feedforward_layernorm": false, "model.encoder.vision_tower.encoder.layers.2.self_attn.k_norm": false, "model.encoder.vision_tower.encoder.layers.2.self_attn.k_proj.linear": false, "model.encoder.vision_tower.encoder.layers.2.self_attn.o_proj.linear": false, "model.encoder.vision_tower.encoder.layers.2.self_attn.q_norm": false, "model.encoder.vision_tower.encoder.layers.2.self_attn.q_proj.linear": false, "model.encoder.vision_tower.encoder.layers.2.self_attn.v_proj.linear": false, "model.encoder.vision_tower.encoder.layers.20.input_layernorm": false, "model.encoder.vision_tower.encoder.layers.20.mlp.down_proj.linear": false, "model.encoder.vision_tower.encoder.layers.20.mlp.gate_proj.linear": false, "model.encoder.vision_tower.encoder.layers.20.mlp.up_proj.linear": false, "model.encoder.vision_tower.encoder.layers.20.post_attention_layernorm": false, "model.encoder.vision_tower.encoder.layers.20.post_feedforward_layernorm": false, "model.encoder.vision_tower.encoder.layers.20.pre_feedforward_layernorm": false, "model.encoder.vision_tower.encoder.layers.20.self_attn.k_norm": false, "model.encoder.vision_tower.encoder.layers.20.self_attn.k_proj.linear": false, "model.encoder.vision_tower.encoder.layers.20.self_attn.o_proj.linear": false, "model.encoder.vision_tower.encoder.layers.20.self_attn.q_norm": false, "model.encoder.vision_tower.encoder.layers.20.self_attn.q_proj.linear": false, "model.encoder.vision_tower.encoder.layers.20.self_attn.v_proj.linear": false, "model.encoder.vision_tower.encoder.layers.21.input_layernorm": false, "model.encoder.vision_tower.encoder.layers.21.mlp.down_proj.linear": false, "model.encoder.vision_tower.encoder.layers.21.mlp.gate_proj.linear": false, "model.encoder.vision_tower.encoder.layers.21.mlp.up_proj.linear": false, "model.encoder.vision_tower.encoder.layers.21.post_attention_layernorm": false, "model.encoder.vision_tower.encoder.layers.21.post_feedforward_layernorm": false, "model.encoder.vision_tower.encoder.layers.21.pre_feedforward_layernorm": false, "model.encoder.vision_tower.encoder.layers.21.self_attn.k_norm": false, "model.encoder.vision_tower.encoder.layers.21.self_attn.k_proj.linear": false, "model.encoder.vision_tower.encoder.layers.21.self_attn.o_proj.linear": false, "model.encoder.vision_tower.encoder.layers.21.self_attn.q_norm": false, "model.encoder.vision_tower.encoder.layers.21.self_attn.q_proj.linear": false, "model.encoder.vision_tower.encoder.layers.21.self_attn.v_proj.linear": false, "model.encoder.vision_tower.encoder.layers.22.input_layernorm": false, "model.encoder.vision_tower.encoder.layers.22.mlp.down_proj.linear": false, "model.encoder.vision_tower.encoder.layers.22.mlp.gate_proj.linear": false, "model.encoder.vision_tower.encoder.layers.22.mlp.up_proj.linear": false, "model.encoder.vision_tower.encoder.layers.22.post_attention_layernorm": false, "model.encoder.vision_tower.encoder.layers.22.post_feedforward_layernorm": false, "model.encoder.vision_tower.encoder.layers.22.pre_feedforward_layernorm": false, "model.encoder.vision_tower.encoder.layers.22.self_attn.k_norm": false, "model.encoder.vision_tower.encoder.layers.22.self_attn.k_proj.linear": false, "model.encoder.vision_tower.encoder.layers.22.self_attn.o_proj.linear": false, "model.encoder.vision_tower.encoder.layers.22.self_attn.q_norm": false, "model.encoder.vision_tower.encoder.layers.22.self_attn.q_proj.linear": false, "model.encoder.vision_tower.encoder.layers.22.self_attn.v_proj.linear": false, "model.encoder.vision_tower.encoder.layers.23.input_layernorm": false, "model.encoder.vision_tower.encoder.layers.23.mlp.down_proj.linear": false, "model.encoder.vision_tower.encoder.layers.23.mlp.gate_proj.linear": false, "model.encoder.vision_tower.encoder.layers.23.mlp.up_proj.linear": false, "model.encoder.vision_tower.encoder.layers.23.post_attention_layernorm": false, "model.encoder.vision_tower.encoder.layers.23.post_feedforward_layernorm": false, "model.encoder.vision_tower.encoder.layers.23.pre_feedforward_layernorm": false, "model.encoder.vision_tower.encoder.layers.23.self_attn.k_norm": false, "model.encoder.vision_tower.encoder.layers.23.self_attn.k_proj.linear": false, "model.encoder.vision_tower.encoder.layers.23.self_attn.o_proj.linear": false, "model.encoder.vision_tower.encoder.layers.23.self_attn.q_norm": false, "model.encoder.vision_tower.encoder.layers.23.self_attn.q_proj.linear": false, "model.encoder.vision_tower.encoder.layers.23.self_attn.v_proj.linear": false, "model.encoder.vision_tower.encoder.layers.24.input_layernorm": false, "model.encoder.vision_tower.encoder.layers.24.mlp.down_proj.linear": false, "model.encoder.vision_tower.encoder.layers.24.mlp.gate_proj.linear": false, "model.encoder.vision_tower.encoder.layers.24.mlp.up_proj.linear": false, "model.encoder.vision_tower.encoder.layers.24.post_attention_layernorm": false, "model.encoder.vision_tower.encoder.layers.24.post_feedforward_layernorm": false, "model.encoder.vision_tower.encoder.layers.24.pre_feedforward_layernorm": false, "model.encoder.vision_tower.encoder.layers.24.self_attn.k_norm": false, "model.encoder.vision_tower.encoder.layers.24.self_attn.k_proj.linear": false, "model.encoder.vision_tower.encoder.layers.24.self_attn.o_proj.linear": false, "model.encoder.vision_tower.encoder.layers.24.self_attn.q_norm": false, "model.encoder.vision_tower.encoder.layers.24.self_attn.q_proj.linear": false, "model.encoder.vision_tower.encoder.layers.24.self_attn.v_proj.linear": false, "model.encoder.vision_tower.encoder.layers.25.input_layernorm": false, "model.encoder.vision_tower.encoder.layers.25.mlp.down_proj.linear": false, "model.encoder.vision_tower.encoder.layers.25.mlp.gate_proj.linear": false, "model.encoder.vision_tower.encoder.layers.25.mlp.up_proj.linear": false, "model.encoder.vision_tower.encoder.layers.25.post_attention_layernorm": false, "model.encoder.vision_tower.encoder.layers.25.post_feedforward_layernorm": false, "model.encoder.vision_tower.encoder.layers.25.pre_feedforward_layernorm": false, "model.encoder.vision_tower.encoder.layers.25.self_attn.k_norm": false, "model.encoder.vision_tower.encoder.layers.25.self_attn.k_proj.linear": false, "model.encoder.vision_tower.encoder.layers.25.self_attn.o_proj.linear": false, "model.encoder.vision_tower.encoder.layers.25.self_attn.q_norm": false, "model.encoder.vision_tower.encoder.layers.25.self_attn.q_proj.linear": false, "model.encoder.vision_tower.encoder.layers.25.self_attn.v_proj.linear": false, "model.encoder.vision_tower.encoder.layers.26.input_layernorm": false, "model.encoder.vision_tower.encoder.layers.26.mlp.down_proj.linear": false, "model.encoder.vision_tower.encoder.layers.26.mlp.gate_proj.linear": false, "model.encoder.vision_tower.encoder.layers.26.mlp.up_proj.linear": false, "model.encoder.vision_tower.encoder.layers.26.post_attention_layernorm": false, "model.encoder.vision_tower.encoder.layers.26.post_feedforward_layernorm": false, "model.encoder.vision_tower.encoder.layers.26.pre_feedforward_layernorm": false, "model.encoder.vision_tower.encoder.layers.26.self_attn.k_norm": false, "model.encoder.vision_tower.encoder.layers.26.self_attn.k_proj.linear": false, "model.encoder.vision_tower.encoder.layers.26.self_attn.o_proj.linear": false, "model.encoder.vision_tower.encoder.layers.26.self_attn.q_norm": false, "model.encoder.vision_tower.encoder.layers.26.self_attn.q_proj.linear": false, "model.encoder.vision_tower.encoder.layers.26.self_attn.v_proj.linear": false, "model.encoder.vision_tower.encoder.layers.3.input_layernorm": false, "model.encoder.vision_tower.encoder.layers.3.mlp.down_proj.linear": false, "model.encoder.vision_tower.encoder.layers.3.mlp.gate_proj.linear": false, "model.encoder.vision_tower.encoder.layers.3.mlp.up_proj.linear": false, "model.encoder.vision_tower.encoder.layers.3.post_attention_layernorm": false, "model.encoder.vision_tower.encoder.layers.3.post_feedforward_layernorm": false, "model.encoder.vision_tower.encoder.layers.3.pre_feedforward_layernorm": false, "model.encoder.vision_tower.encoder.layers.3.self_attn.k_norm": false, "model.encoder.vision_tower.encoder.layers.3.self_attn.k_proj.linear": false, "model.encoder.vision_tower.encoder.layers.3.self_attn.o_proj.linear": false, "model.encoder.vision_tower.encoder.layers.3.self_attn.q_norm": false, "model.encoder.vision_tower.encoder.layers.3.self_attn.q_proj.linear": false, "model.encoder.vision_tower.encoder.layers.3.self_attn.v_proj.linear": false, "model.encoder.vision_tower.encoder.layers.4.input_layernorm": false, "model.encoder.vision_tower.encoder.layers.4.mlp.down_proj.linear": false, "model.encoder.vision_tower.encoder.layers.4.mlp.gate_proj.linear": false, "model.encoder.vision_tower.encoder.layers.4.mlp.up_proj.linear": false, "model.encoder.vision_tower.encoder.layers.4.post_attention_layernorm": false, "model.encoder.vision_tower.encoder.layers.4.post_feedforward_layernorm": false, "model.encoder.vision_tower.encoder.layers.4.pre_feedforward_layernorm": false, "model.encoder.vision_tower.encoder.layers.4.self_attn.k_norm": false, "model.encoder.vision_tower.encoder.layers.4.self_attn.k_proj.linear": false, "model.encoder.vision_tower.encoder.layers.4.self_attn.o_proj.linear": false, "model.encoder.vision_tower.encoder.layers.4.self_attn.q_norm": false, "model.encoder.vision_tower.encoder.layers.4.self_attn.q_proj.linear": false, "model.encoder.vision_tower.encoder.layers.4.self_attn.v_proj.linear": false, "model.encoder.vision_tower.encoder.layers.5.input_layernorm": false, "model.encoder.vision_tower.encoder.layers.5.mlp.down_proj.linear": false, "model.encoder.vision_tower.encoder.layers.5.mlp.gate_proj.linear": false, "model.encoder.vision_tower.encoder.layers.5.mlp.up_proj.linear": false, "model.encoder.vision_tower.encoder.layers.5.post_attention_layernorm": false, "model.encoder.vision_tower.encoder.layers.5.post_feedforward_layernorm": false, "model.encoder.vision_tower.encoder.layers.5.pre_feedforward_layernorm": false, "model.encoder.vision_tower.encoder.layers.5.self_attn.k_norm": false, "model.encoder.vision_tower.encoder.layers.5.self_attn.k_proj.linear": false, "model.encoder.vision_tower.encoder.layers.5.self_attn.o_proj.linear": false, "model.encoder.vision_tower.encoder.layers.5.self_attn.q_norm": false, "model.encoder.vision_tower.encoder.layers.5.self_attn.q_proj.linear": false, "model.encoder.vision_tower.encoder.layers.5.self_attn.v_proj.linear": false, "model.encoder.vision_tower.encoder.layers.6.input_layernorm": false, "model.encoder.vision_tower.encoder.layers.6.mlp.down_proj.linear": false, "model.encoder.vision_tower.encoder.layers.6.mlp.gate_proj.linear": false, "model.encoder.vision_tower.encoder.layers.6.mlp.up_proj.linear": false, "model.encoder.vision_tower.encoder.layers.6.post_attention_layernorm": false, "model.encoder.vision_tower.encoder.layers.6.post_feedforward_layernorm": false, "model.encoder.vision_tower.encoder.layers.6.pre_feedforward_layernorm": false, "model.encoder.vision_tower.encoder.layers.6.self_attn.k_norm": false, "model.encoder.vision_tower.encoder.layers.6.self_attn.k_proj.linear": false, "model.encoder.vision_tower.encoder.layers.6.self_attn.o_proj.linear": false, "model.encoder.vision_tower.encoder.layers.6.self_attn.q_norm": false, "model.encoder.vision_tower.encoder.layers.6.self_attn.q_proj.linear": false, "model.encoder.vision_tower.encoder.layers.6.self_attn.v_proj.linear": false, "model.encoder.vision_tower.encoder.layers.7.input_layernorm": false, "model.encoder.vision_tower.encoder.layers.7.mlp.down_proj.linear": false, "model.encoder.vision_tower.encoder.layers.7.mlp.gate_proj.linear": false, "model.encoder.vision_tower.encoder.layers.7.mlp.up_proj.linear": false, "model.encoder.vision_tower.encoder.layers.7.post_attention_layernorm": false, "model.encoder.vision_tower.encoder.layers.7.post_feedforward_layernorm": false, "model.encoder.vision_tower.encoder.layers.7.pre_feedforward_layernorm": false, "model.encoder.vision_tower.encoder.layers.7.self_attn.k_norm": false, "model.encoder.vision_tower.encoder.layers.7.self_attn.k_proj.linear": false, "model.encoder.vision_tower.encoder.layers.7.self_attn.o_proj.linear": false, "model.encoder.vision_tower.encoder.layers.7.self_attn.q_norm": false, "model.encoder.vision_tower.encoder.layers.7.self_attn.q_proj.linear": false, "model.encoder.vision_tower.encoder.layers.7.self_attn.v_proj.linear": false, "model.encoder.vision_tower.encoder.layers.8.input_layernorm": false, "model.encoder.vision_tower.encoder.layers.8.mlp.down_proj.linear": false, "model.encoder.vision_tower.encoder.layers.8.mlp.gate_proj.linear": false, "model.encoder.vision_tower.encoder.layers.8.mlp.up_proj.linear": false, "model.encoder.vision_tower.encoder.layers.8.post_attention_layernorm": false, "model.encoder.vision_tower.encoder.layers.8.post_feedforward_layernorm": false, "model.encoder.vision_tower.encoder.layers.8.pre_feedforward_layernorm": false, "model.encoder.vision_tower.encoder.layers.8.self_attn.k_norm": false, "model.encoder.vision_tower.encoder.layers.8.self_attn.k_proj.linear": false, "model.encoder.vision_tower.encoder.layers.8.self_attn.o_proj.linear": false, "model.encoder.vision_tower.encoder.layers.8.self_attn.q_norm": false, "model.encoder.vision_tower.encoder.layers.8.self_attn.q_proj.linear": false, "model.encoder.vision_tower.encoder.layers.8.self_attn.v_proj.linear": false, "model.encoder.vision_tower.encoder.layers.9.input_layernorm": false, "model.encoder.vision_tower.encoder.layers.9.mlp.down_proj.linear": false, "model.encoder.vision_tower.encoder.layers.9.mlp.gate_proj.linear": false, "model.encoder.vision_tower.encoder.layers.9.mlp.up_proj.linear": false, "model.encoder.vision_tower.encoder.layers.9.post_attention_layernorm": false, "model.encoder.vision_tower.encoder.layers.9.post_feedforward_layernorm": false, "model.encoder.vision_tower.encoder.layers.9.pre_feedforward_layernorm": false, "model.encoder.vision_tower.encoder.layers.9.self_attn.k_norm": false, "model.encoder.vision_tower.encoder.layers.9.self_attn.k_proj.linear": false, "model.encoder.vision_tower.encoder.layers.9.self_attn.o_proj.linear": false, "model.encoder.vision_tower.encoder.layers.9.self_attn.q_norm": false, "model.encoder.vision_tower.encoder.layers.9.self_attn.q_proj.linear": false, "model.encoder.vision_tower.encoder.layers.9.self_attn.v_proj.linear": false, "model.encoder.vision_tower.patch_embedder.input_proj": false, "passthrough_roles": [ "embedding", "norms", "router_scalars", "self_conditioning", "vision_encoder" ], "quantization_backend": "mx.quantize", "quantized_roles": [ "attention", "dense_mlp", "experts", "router" ] }, "text_config": { "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 2, "dtype": "bfloat16", "eos_token_id": 1, "final_logit_softcapping": 30.0, "global_head_dim": 512, "head_dim": 256, "hidden_activation": "gelu_pytorch_tanh", "hidden_size": 2816, "initializer_range": 0.02, "intermediate_size": 2112, "layer_types": [ "sliding_attention", "sliding_attention", "sliding_attention", "sliding_attention", "sliding_attention", "full_attention", "sliding_attention", "sliding_attention", "sliding_attention", "sliding_attention", "sliding_attention", "full_attention", "sliding_attention", "sliding_attention", "sliding_attention", "sliding_attention", "sliding_attention", "full_attention", "sliding_attention", "sliding_attention", "sliding_attention", "sliding_attention", "sliding_attention", "full_attention", "sliding_attention", "sliding_attention", "sliding_attention", "sliding_attention", "sliding_attention", "full_attention" ], "max_position_embeddings": 262144, "model_type": "diffusion_gemma_text", "moe_intermediate_size": 704, "num_attention_heads": 16, "num_experts": 128, "num_global_key_value_heads": 2, "num_hidden_layers": 30, "num_key_value_heads": 8, "pad_token_id": 0, "rms_norm_eps": 1e-06, "rope_parameters": { "full_attention": { "partial_rotary_factor": 0.25, "rope_theta": 1000000.0, "rope_type": "proportional" }, "sliding_attention": { "rope_theta": 10000.0, "rope_type": "default" } }, "sliding_window": 1024, "tie_word_embeddings": true, "top_k_experts": 8, "use_bidirectional_attention": "vision", "vocab_size": 262144 }, "tie_word_embeddings": true, "transformers_version": "5.8.0.dev0", "vision_config": { "_name_or_path": "", "architectures": null, "attention_bias": false, "attention_dropout": 0.0, "chunk_size_feed_forward": 0, "default_output_length": 280, "dtype": "bfloat16", "global_head_dim": 72, "head_dim": 72, "hidden_activation": "gelu_pytorch_tanh", "hidden_size": 1152, "id2label": { "0": "LABEL_0", "1": "LABEL_1" }, "initializer_range": 0.02, "intermediate_size": 4304, "is_encoder_decoder": false, "label2id": { "LABEL_0": 0, "LABEL_1": 1 }, "max_position_embeddings": 131072, "model_type": "gemma4_vision", "num_attention_heads": 16, "num_hidden_layers": 27, "num_key_value_heads": 16, "output_attentions": false, "output_hidden_states": false, "patch_size": 16, "pooling_kernel_size": 3, "position_embedding_size": 10240, "problem_type": null, "return_dict": true, "rms_norm_eps": 1e-06, "rope_parameters": { "rope_theta": 100.0, "rope_type": "default" }, "standardize": true, "use_clipped_linears": false }, "vision_soft_tokens_per_image": 280, "weight_format": "mxfp8" }