| { |
| "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_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_group_norm_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", |
| "quantization_config": { |
| "config_groups": { |
| "group_0": { |
| "input_activations": null, |
| "output_activations": null, |
| "targets": [ |
| "Linear" |
| ], |
| "weights": { |
| "actorder": null, |
| "block_structure": null, |
| "dynamic": false, |
| "group_size": 128, |
| "num_bits": 4, |
| "observer": "minmax", |
| "observer_kwargs": {}, |
| "strategy": "group", |
| "symmetric": true, |
| "type": "int" |
| } |
| } |
| }, |
| "format": "pack-quantized", |
| "global_compression_ratio": null, |
| "ignore": [ |
| "model.language_model.layers.0.altup.correction_coefs", |
| "model.language_model.layers.0.altup.prediction_coefs", |
| "model.language_model.layers.0.altup.modality_router", |
| "model.language_model.layers.0.laurel.linear_left", |
| "model.language_model.layers.0.laurel.linear_right", |
| "model.language_model.layers.0.per_layer_input_gate", |
| "model.language_model.layers.0.per_layer_projection", |
| "model.language_model.layers.1.altup.correction_coefs", |
| "model.language_model.layers.1.altup.prediction_coefs", |
| "model.language_model.layers.1.altup.modality_router", |
| "model.language_model.layers.1.laurel.linear_left", |
| "model.language_model.layers.1.laurel.linear_right", |
| "model.language_model.layers.1.per_layer_input_gate", |
| "model.language_model.layers.1.per_layer_projection", |
| "model.language_model.layers.2.altup.correction_coefs", |
| "model.language_model.layers.2.altup.prediction_coefs", |
| "model.language_model.layers.2.altup.modality_router", |
| "model.language_model.layers.2.laurel.linear_left", |
| "model.language_model.layers.2.laurel.linear_right", |
| "model.language_model.layers.2.per_layer_input_gate", |
| "model.language_model.layers.2.per_layer_projection", |
| "model.language_model.layers.3.altup.correction_coefs", |
| "model.language_model.layers.3.altup.prediction_coefs", |
| "model.language_model.layers.3.altup.modality_router", |
| "model.language_model.layers.3.laurel.linear_left", |
| "model.language_model.layers.3.laurel.linear_right", |
| "model.language_model.layers.3.per_layer_input_gate", |
| "model.language_model.layers.3.per_layer_projection", |
| "model.language_model.layers.4.altup.correction_coefs", |
| "model.language_model.layers.4.altup.prediction_coefs", |
| "model.language_model.layers.4.altup.modality_router", |
| "model.language_model.layers.4.laurel.linear_left", |
| "model.language_model.layers.4.laurel.linear_right", |
| "model.language_model.layers.4.per_layer_input_gate", |
| "model.language_model.layers.4.per_layer_projection", |
| "model.language_model.layers.5.altup.correction_coefs", |
| "model.language_model.layers.5.altup.prediction_coefs", |
| "model.language_model.layers.5.altup.modality_router", |
| "model.language_model.layers.5.laurel.linear_left", |
| "model.language_model.layers.5.laurel.linear_right", |
| "model.language_model.layers.5.per_layer_input_gate", |
| "model.language_model.layers.5.per_layer_projection", |
| "model.language_model.layers.6.altup.correction_coefs", |
| "model.language_model.layers.6.altup.prediction_coefs", |
| "model.language_model.layers.6.altup.modality_router", |
| "model.language_model.layers.6.laurel.linear_left", |
| "model.language_model.layers.6.laurel.linear_right", |
| "model.language_model.layers.6.per_layer_input_gate", |
| "model.language_model.layers.6.per_layer_projection", |
| "model.language_model.layers.7.altup.correction_coefs", |
| "model.language_model.layers.7.altup.prediction_coefs", |
| "model.language_model.layers.7.altup.modality_router", |
| "model.language_model.layers.7.laurel.linear_left", |
| "model.language_model.layers.7.laurel.linear_right", |
| "model.language_model.layers.7.per_layer_input_gate", |
| "model.language_model.layers.7.per_layer_projection", |
| "model.language_model.layers.8.altup.correction_coefs", |
| "model.language_model.layers.8.altup.prediction_coefs", |
| "model.language_model.layers.8.altup.modality_router", |
| "model.language_model.layers.8.laurel.linear_left", |
| "model.language_model.layers.8.laurel.linear_right", |
| "model.language_model.layers.8.per_layer_input_gate", |
| "model.language_model.layers.8.per_layer_projection", |
| "model.language_model.layers.9.altup.correction_coefs", |
| "model.language_model.layers.9.altup.prediction_coefs", |
| "model.language_model.layers.9.altup.modality_router", |
| "model.language_model.layers.9.laurel.linear_left", |
| "model.language_model.layers.9.laurel.linear_right", |
| "model.language_model.layers.9.per_layer_input_gate", |
| "model.language_model.layers.9.per_layer_projection", |
| "model.language_model.layers.10.altup.correction_coefs", |
| "model.language_model.layers.10.altup.prediction_coefs", |
| "model.language_model.layers.10.altup.modality_router", |
| "model.language_model.layers.10.laurel.linear_left", |
| "model.language_model.layers.10.laurel.linear_right", |
| "model.language_model.layers.10.per_layer_input_gate", |
| "model.language_model.layers.10.per_layer_projection", |
| "model.language_model.layers.11.altup.correction_coefs", |
| "model.language_model.layers.11.altup.prediction_coefs", |
| "model.language_model.layers.11.altup.modality_router", |
| "model.language_model.layers.11.laurel.linear_left", |
| "model.language_model.layers.11.laurel.linear_right", |
| "model.language_model.layers.11.per_layer_input_gate", |
| "model.language_model.layers.11.per_layer_projection", |
| "model.language_model.layers.12.altup.correction_coefs", |
| "model.language_model.layers.12.altup.prediction_coefs", |
| "model.language_model.layers.12.altup.modality_router", |
| "model.language_model.layers.12.laurel.linear_left", |
| "model.language_model.layers.12.laurel.linear_right", |
| "model.language_model.layers.12.per_layer_input_gate", |
| "model.language_model.layers.12.per_layer_projection", |
| "model.language_model.layers.13.altup.correction_coefs", |
| "model.language_model.layers.13.altup.prediction_coefs", |
| "model.language_model.layers.13.altup.modality_router", |
| "model.language_model.layers.13.laurel.linear_left", |
| "model.language_model.layers.13.laurel.linear_right", |
| "model.language_model.layers.13.per_layer_input_gate", |
| "model.language_model.layers.13.per_layer_projection", |
| "model.language_model.layers.14.altup.correction_coefs", |
| "model.language_model.layers.14.altup.prediction_coefs", |
| "model.language_model.layers.14.altup.modality_router", |
| "model.language_model.layers.14.laurel.linear_left", |
| "model.language_model.layers.14.laurel.linear_right", |
| "model.language_model.layers.14.per_layer_input_gate", |
| "model.language_model.layers.14.per_layer_projection", |
| "model.language_model.layers.15.altup.correction_coefs", |
| "model.language_model.layers.15.altup.prediction_coefs", |
| "model.language_model.layers.15.altup.modality_router", |
| "model.language_model.layers.15.laurel.linear_left", |
| "model.language_model.layers.15.laurel.linear_right", |
| "model.language_model.layers.15.per_layer_input_gate", |
| "model.language_model.layers.15.per_layer_projection", |
| "model.language_model.layers.16.altup.correction_coefs", |
| "model.language_model.layers.16.altup.prediction_coefs", |
| "model.language_model.layers.16.altup.modality_router", |
| "model.language_model.layers.16.laurel.linear_left", |
| "model.language_model.layers.16.laurel.linear_right", |
| "model.language_model.layers.16.per_layer_input_gate", |
| "model.language_model.layers.16.per_layer_projection", |
| "model.language_model.layers.17.altup.correction_coefs", |
| "model.language_model.layers.17.altup.prediction_coefs", |
| "model.language_model.layers.17.altup.modality_router", |
| "model.language_model.layers.17.laurel.linear_left", |
| "model.language_model.layers.17.laurel.linear_right", |
| "model.language_model.layers.17.per_layer_input_gate", |
| "model.language_model.layers.17.per_layer_projection", |
| "model.language_model.layers.18.altup.correction_coefs", |
| "model.language_model.layers.18.altup.prediction_coefs", |
| "model.language_model.layers.18.altup.modality_router", |
| "model.language_model.layers.18.laurel.linear_left", |
| "model.language_model.layers.18.laurel.linear_right", |
| "model.language_model.layers.18.per_layer_input_gate", |
| "model.language_model.layers.18.per_layer_projection", |
| "model.language_model.layers.19.altup.correction_coefs", |
| "model.language_model.layers.19.altup.prediction_coefs", |
| "model.language_model.layers.19.altup.modality_router", |
| "model.language_model.layers.19.laurel.linear_left", |
| "model.language_model.layers.19.laurel.linear_right", |
| "model.language_model.layers.19.per_layer_input_gate", |
| "model.language_model.layers.19.per_layer_projection", |
| "model.language_model.layers.20.altup.correction_coefs", |
| "model.language_model.layers.20.altup.prediction_coefs", |
| "model.language_model.layers.20.altup.modality_router", |
| "model.language_model.layers.20.laurel.linear_left", |
| "model.language_model.layers.20.laurel.linear_right", |
| "model.language_model.layers.20.per_layer_input_gate", |
| "model.language_model.layers.20.per_layer_projection", |
| "model.language_model.layers.21.altup.correction_coefs", |
| "model.language_model.layers.21.altup.prediction_coefs", |
| "model.language_model.layers.21.altup.modality_router", |
| "model.language_model.layers.21.laurel.linear_left", |
| "model.language_model.layers.21.laurel.linear_right", |
| "model.language_model.layers.21.per_layer_input_gate", |
| "model.language_model.layers.21.per_layer_projection", |
| "model.language_model.layers.22.altup.correction_coefs", |
| "model.language_model.layers.22.altup.prediction_coefs", |
| "model.language_model.layers.22.altup.modality_router", |
| "model.language_model.layers.22.laurel.linear_left", |
| "model.language_model.layers.22.laurel.linear_right", |
| "model.language_model.layers.22.per_layer_input_gate", |
| "model.language_model.layers.22.per_layer_projection", |
| "model.language_model.layers.23.altup.correction_coefs", |
| "model.language_model.layers.23.altup.prediction_coefs", |
| "model.language_model.layers.23.altup.modality_router", |
| "model.language_model.layers.23.laurel.linear_left", |
| "model.language_model.layers.23.laurel.linear_right", |
| "model.language_model.layers.23.per_layer_input_gate", |
| "model.language_model.layers.23.per_layer_projection", |
| "model.language_model.layers.24.altup.correction_coefs", |
| "model.language_model.layers.24.altup.prediction_coefs", |
| "model.language_model.layers.24.altup.modality_router", |
| "model.language_model.layers.24.laurel.linear_left", |
| "model.language_model.layers.24.laurel.linear_right", |
| "model.language_model.layers.24.per_layer_input_gate", |
| "model.language_model.layers.24.per_layer_projection", |
| "model.language_model.layers.25.altup.correction_coefs", |
| "model.language_model.layers.25.altup.prediction_coefs", |
| "model.language_model.layers.25.altup.modality_router", |
| "model.language_model.layers.25.laurel.linear_left", |
| "model.language_model.layers.25.laurel.linear_right", |
| "model.language_model.layers.25.per_layer_input_gate", |
| "model.language_model.layers.25.per_layer_projection", |
| "model.language_model.layers.26.altup.correction_coefs", |
| "model.language_model.layers.26.altup.prediction_coefs", |
| "model.language_model.layers.26.altup.modality_router", |
| "model.language_model.layers.26.laurel.linear_left", |
| "model.language_model.layers.26.laurel.linear_right", |
| "model.language_model.layers.26.per_layer_input_gate", |
| "model.language_model.layers.26.per_layer_projection", |
| "model.language_model.layers.27.altup.correction_coefs", |
| "model.language_model.layers.27.altup.prediction_coefs", |
| "model.language_model.layers.27.altup.modality_router", |
| "model.language_model.layers.27.laurel.linear_left", |
| "model.language_model.layers.27.laurel.linear_right", |
| "model.language_model.layers.27.per_layer_input_gate", |
| "model.language_model.layers.27.per_layer_projection", |
| "model.language_model.layers.28.altup.correction_coefs", |
| "model.language_model.layers.28.altup.prediction_coefs", |
| "model.language_model.layers.28.altup.modality_router", |
| "model.language_model.layers.28.laurel.linear_left", |
| "model.language_model.layers.28.laurel.linear_right", |
| "model.language_model.layers.28.per_layer_input_gate", |
| "model.language_model.layers.28.per_layer_projection", |
| "model.language_model.layers.29.altup.correction_coefs", |
| "model.language_model.layers.29.altup.prediction_coefs", |
| "model.language_model.layers.29.altup.modality_router", |
| "model.language_model.layers.29.laurel.linear_left", |
| "model.language_model.layers.29.laurel.linear_right", |
| "model.language_model.layers.29.per_layer_input_gate", |
| "model.language_model.layers.29.per_layer_projection", |
| "model.language_model.layers.30.altup.correction_coefs", |
| "model.language_model.layers.30.altup.prediction_coefs", |
| "model.language_model.layers.30.altup.modality_router", |
| "model.language_model.layers.30.laurel.linear_left", |
| "model.language_model.layers.30.laurel.linear_right", |
| "model.language_model.layers.30.per_layer_input_gate", |
| "model.language_model.layers.30.per_layer_projection", |
| "model.language_model.layers.31.altup.correction_coefs", |
| "model.language_model.layers.31.altup.prediction_coefs", |
| "model.language_model.layers.31.altup.modality_router", |
| "model.language_model.layers.31.laurel.linear_left", |
| "model.language_model.layers.31.laurel.linear_right", |
| "model.language_model.layers.31.per_layer_input_gate", |
| "model.language_model.layers.31.per_layer_projection", |
| "model.language_model.layers.32.altup.correction_coefs", |
| "model.language_model.layers.32.altup.prediction_coefs", |
| "model.language_model.layers.32.altup.modality_router", |
| "model.language_model.layers.32.laurel.linear_left", |
| "model.language_model.layers.32.laurel.linear_right", |
| "model.language_model.layers.32.per_layer_input_gate", |
| "model.language_model.layers.32.per_layer_projection", |
| "model.language_model.layers.33.altup.correction_coefs", |
| "model.language_model.layers.33.altup.prediction_coefs", |
| "model.language_model.layers.33.altup.modality_router", |
| "model.language_model.layers.33.laurel.linear_left", |
| "model.language_model.layers.33.laurel.linear_right", |
| "model.language_model.layers.33.per_layer_input_gate", |
| "model.language_model.layers.33.per_layer_projection", |
| "model.language_model.layers.34.altup.correction_coefs", |
| "model.language_model.layers.34.altup.prediction_coefs", |
| "model.language_model.layers.34.altup.modality_router", |
| "model.language_model.layers.34.laurel.linear_left", |
| "model.language_model.layers.34.laurel.linear_right", |
| "model.language_model.layers.34.per_layer_input_gate", |
| "model.language_model.layers.34.per_layer_projection", |
| "model.language_model.per_layer_model_projection", |
| "model.language_model.altup_projections.0", |
| "model.language_model.altup_projections.1", |
| "model.language_model.altup_projections.2", |
| "model.language_model.altup_unembed_projections.0", |
| "model.language_model.altup_unembed_projections.1", |
| "model.language_model.altup_unembed_projections.2", |
| "model.audio_tower.subsample_conv_projection.input_proj_linear", |
| "model.audio_tower.conformer.0.ffw_layer_start.ffw_layer_1", |
| "model.audio_tower.conformer.0.ffw_layer_start.ffw_layer_2", |
| "model.audio_tower.conformer.0.attention.attn.relative_position_embedding.pos_proj", |
| "model.audio_tower.conformer.0.attention.attn.q_proj", |
| "model.audio_tower.conformer.0.attention.attn.k_proj", |
| "model.audio_tower.conformer.0.attention.attn.v_proj", |
| "model.audio_tower.conformer.0.attention.post", |
| "model.audio_tower.conformer.0.lconv1d.linear_start", |
| "model.audio_tower.conformer.0.lconv1d.linear_end", |
| "model.audio_tower.conformer.0.ffw_layer_end.ffw_layer_1", |
| "model.audio_tower.conformer.0.ffw_layer_end.ffw_layer_2", |
| "model.audio_tower.conformer.1.ffw_layer_start.ffw_layer_1", |
| "model.audio_tower.conformer.1.ffw_layer_start.ffw_layer_2", |
| "model.audio_tower.conformer.1.attention.attn.relative_position_embedding.pos_proj", |
| "model.audio_tower.conformer.1.attention.attn.q_proj", |
| "model.audio_tower.conformer.1.attention.attn.k_proj", |
| "model.audio_tower.conformer.1.attention.attn.v_proj", |
| "model.audio_tower.conformer.1.attention.post", |
| "model.audio_tower.conformer.1.lconv1d.linear_start", |
| "model.audio_tower.conformer.1.lconv1d.linear_end", |
| "model.audio_tower.conformer.1.ffw_layer_end.ffw_layer_1", |
| "model.audio_tower.conformer.1.ffw_layer_end.ffw_layer_2", |
| "model.audio_tower.conformer.2.ffw_layer_start.ffw_layer_1", |
| "model.audio_tower.conformer.2.ffw_layer_start.ffw_layer_2", |
| "model.audio_tower.conformer.2.attention.attn.relative_position_embedding.pos_proj", |
| "model.audio_tower.conformer.2.attention.attn.q_proj", |
| "model.audio_tower.conformer.2.attention.attn.k_proj", |
| "model.audio_tower.conformer.2.attention.attn.v_proj", |
| "model.audio_tower.conformer.2.attention.post", |
| "model.audio_tower.conformer.2.lconv1d.linear_start", |
| "model.audio_tower.conformer.2.lconv1d.linear_end", |
| "model.audio_tower.conformer.2.ffw_layer_end.ffw_layer_1", |
| "model.audio_tower.conformer.2.ffw_layer_end.ffw_layer_2", |
| "model.audio_tower.conformer.3.ffw_layer_start.ffw_layer_1", |
| "model.audio_tower.conformer.3.ffw_layer_start.ffw_layer_2", |
| "model.audio_tower.conformer.3.attention.attn.relative_position_embedding.pos_proj", |
| "model.audio_tower.conformer.3.attention.attn.q_proj", |
| "model.audio_tower.conformer.3.attention.attn.k_proj", |
| "model.audio_tower.conformer.3.attention.attn.v_proj", |
| "model.audio_tower.conformer.3.attention.post", |
| "model.audio_tower.conformer.3.lconv1d.linear_start", |
| "model.audio_tower.conformer.3.lconv1d.linear_end", |
| "model.audio_tower.conformer.3.ffw_layer_end.ffw_layer_1", |
| "model.audio_tower.conformer.3.ffw_layer_end.ffw_layer_2", |
| "model.audio_tower.conformer.4.ffw_layer_start.ffw_layer_1", |
| "model.audio_tower.conformer.4.ffw_layer_start.ffw_layer_2", |
| "model.audio_tower.conformer.4.attention.attn.relative_position_embedding.pos_proj", |
| "model.audio_tower.conformer.4.attention.attn.q_proj", |
| "model.audio_tower.conformer.4.attention.attn.k_proj", |
| "model.audio_tower.conformer.4.attention.attn.v_proj", |
| "model.audio_tower.conformer.4.attention.post", |
| "model.audio_tower.conformer.4.lconv1d.linear_start", |
| "model.audio_tower.conformer.4.lconv1d.linear_end", |
| "model.audio_tower.conformer.4.ffw_layer_end.ffw_layer_1", |
| "model.audio_tower.conformer.4.ffw_layer_end.ffw_layer_2", |
| "model.audio_tower.conformer.5.ffw_layer_start.ffw_layer_1", |
| "model.audio_tower.conformer.5.ffw_layer_start.ffw_layer_2", |
| "model.audio_tower.conformer.5.attention.attn.relative_position_embedding.pos_proj", |
| "model.audio_tower.conformer.5.attention.attn.q_proj", |
| "model.audio_tower.conformer.5.attention.attn.k_proj", |
| "model.audio_tower.conformer.5.attention.attn.v_proj", |
| "model.audio_tower.conformer.5.attention.post", |
| "model.audio_tower.conformer.5.lconv1d.linear_start", |
| "model.audio_tower.conformer.5.lconv1d.linear_end", |
| "model.audio_tower.conformer.5.ffw_layer_end.ffw_layer_1", |
| "model.audio_tower.conformer.5.ffw_layer_end.ffw_layer_2", |
| "model.audio_tower.conformer.6.ffw_layer_start.ffw_layer_1", |
| "model.audio_tower.conformer.6.ffw_layer_start.ffw_layer_2", |
| "model.audio_tower.conformer.6.attention.attn.relative_position_embedding.pos_proj", |
| "model.audio_tower.conformer.6.attention.attn.q_proj", |
| "model.audio_tower.conformer.6.attention.attn.k_proj", |
| "model.audio_tower.conformer.6.attention.attn.v_proj", |
| "model.audio_tower.conformer.6.attention.post", |
| "model.audio_tower.conformer.6.lconv1d.linear_start", |
| "model.audio_tower.conformer.6.lconv1d.linear_end", |
| "model.audio_tower.conformer.6.ffw_layer_end.ffw_layer_1", |
| "model.audio_tower.conformer.6.ffw_layer_end.ffw_layer_2", |
| "model.audio_tower.conformer.7.ffw_layer_start.ffw_layer_1", |
| "model.audio_tower.conformer.7.ffw_layer_start.ffw_layer_2", |
| "model.audio_tower.conformer.7.attention.attn.relative_position_embedding.pos_proj", |
| "model.audio_tower.conformer.7.attention.attn.q_proj", |
| "model.audio_tower.conformer.7.attention.attn.k_proj", |
| "model.audio_tower.conformer.7.attention.attn.v_proj", |
| "model.audio_tower.conformer.7.attention.post", |
| "model.audio_tower.conformer.7.lconv1d.linear_start", |
| "model.audio_tower.conformer.7.lconv1d.linear_end", |
| "model.audio_tower.conformer.7.ffw_layer_end.ffw_layer_1", |
| "model.audio_tower.conformer.7.ffw_layer_end.ffw_layer_2", |
| "model.audio_tower.conformer.8.ffw_layer_start.ffw_layer_1", |
| "model.audio_tower.conformer.8.ffw_layer_start.ffw_layer_2", |
| "model.audio_tower.conformer.8.attention.attn.relative_position_embedding.pos_proj", |
| "model.audio_tower.conformer.8.attention.attn.q_proj", |
| "model.audio_tower.conformer.8.attention.attn.k_proj", |
| "model.audio_tower.conformer.8.attention.attn.v_proj", |
| "model.audio_tower.conformer.8.attention.post", |
| "model.audio_tower.conformer.8.lconv1d.linear_start", |
| "model.audio_tower.conformer.8.lconv1d.linear_end", |
| "model.audio_tower.conformer.8.ffw_layer_end.ffw_layer_1", |
| "model.audio_tower.conformer.8.ffw_layer_end.ffw_layer_2", |
| "model.audio_tower.conformer.9.ffw_layer_start.ffw_layer_1", |
| "model.audio_tower.conformer.9.ffw_layer_start.ffw_layer_2", |
| "model.audio_tower.conformer.9.attention.attn.relative_position_embedding.pos_proj", |
| "model.audio_tower.conformer.9.attention.attn.q_proj", |
| "model.audio_tower.conformer.9.attention.attn.k_proj", |
| "model.audio_tower.conformer.9.attention.attn.v_proj", |
| "model.audio_tower.conformer.9.attention.post", |
| "model.audio_tower.conformer.9.lconv1d.linear_start", |
| "model.audio_tower.conformer.9.lconv1d.linear_end", |
| "model.audio_tower.conformer.9.ffw_layer_end.ffw_layer_1", |
| "model.audio_tower.conformer.9.ffw_layer_end.ffw_layer_2", |
| "model.audio_tower.conformer.10.ffw_layer_start.ffw_layer_1", |
| "model.audio_tower.conformer.10.ffw_layer_start.ffw_layer_2", |
| "model.audio_tower.conformer.10.attention.attn.relative_position_embedding.pos_proj", |
| "model.audio_tower.conformer.10.attention.attn.q_proj", |
| "model.audio_tower.conformer.10.attention.attn.k_proj", |
| "model.audio_tower.conformer.10.attention.attn.v_proj", |
| "model.audio_tower.conformer.10.attention.post", |
| "model.audio_tower.conformer.10.lconv1d.linear_start", |
| "model.audio_tower.conformer.10.lconv1d.linear_end", |
| "model.audio_tower.conformer.10.ffw_layer_end.ffw_layer_1", |
| "model.audio_tower.conformer.10.ffw_layer_end.ffw_layer_2", |
| "model.audio_tower.conformer.11.ffw_layer_start.ffw_layer_1", |
| "model.audio_tower.conformer.11.ffw_layer_start.ffw_layer_2", |
| "model.audio_tower.conformer.11.attention.attn.relative_position_embedding.pos_proj", |
| "model.audio_tower.conformer.11.attention.attn.q_proj", |
| "model.audio_tower.conformer.11.attention.attn.k_proj", |
| "model.audio_tower.conformer.11.attention.attn.v_proj", |
| "model.audio_tower.conformer.11.attention.post", |
| "model.audio_tower.conformer.11.lconv1d.linear_start", |
| "model.audio_tower.conformer.11.lconv1d.linear_end", |
| "model.audio_tower.conformer.11.ffw_layer_end.ffw_layer_1", |
| "model.audio_tower.conformer.11.ffw_layer_end.ffw_layer_2", |
| "model.embed_vision.embedding_projection", |
| "model.embed_audio.embedding_projection", |
| "lm_head" |
| ], |
| "kv_cache_scheme": null, |
| "quant_method": "compressed-tensors", |
| "quantization_status": "compressed" |
| }, |
| "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_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, |
| 16384, |
| 16384, |
| 16384, |
| 16384, |
| 16384, |
| 16384, |
| 16384, |
| 16384, |
| 16384, |
| 16384, |
| 16384, |
| 16384, |
| 16384, |
| 16384, |
| 16384, |
| 16384, |
| 16384, |
| 16384, |
| 16384, |
| 16384, |
| 16384, |
| 16384, |
| 16384, |
| 16384, |
| 16384, |
| 16384, |
| 16384, |
| 16384, |
| 16384, |
| 16384, |
| 16384, |
| 16384, |
| 16384, |
| 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, |
| "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 |
| }, |
| "torch_dtype": "bfloat16", |
| "transformers_version": "4.53.2", |
| "vision_config": { |
| "architecture": "mobilenetv5_300m_enc", |
| "do_pooling": false, |
| "hidden_size": 2048, |
| "initializer_range": 0.02, |
| "label_names": [ |
| "LABEL_0", |
| "LABEL_1" |
| ], |
| "model_args": null, |
| "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 |
| } |
