| { |
| "model_name": "models--google--gemma-4-E2B-it", |
| "model_type": "vlm-gemma4", |
| "vm_cfg": { |
| "model_type": "gemma4_vision", |
| "arch": "gemma4_vision", |
| "image_size": [ |
| 480, |
| 480 |
| ], |
| "patch_size": 16, |
| "cls_embed": false, |
| "hidden_size": 768, |
| "intermediate_size": 3072, |
| "num_attention_heads": 12, |
| "num_hidden_layers": 16, |
| "hidden_act": "gelu_pytorch_tanh", |
| "layer_norm_eps": 1e-06, |
| "spatial_merge_size": 3, |
| "temporal_patch_size": 1, |
| "window_size": 0, |
| "num_position_embeddings": 0, |
| "fullatt_block_indexes": [], |
| "deepstack_visual_indexes": [], |
| "rope_theta": 100.0 |
| }, |
| "mm_cfg": { |
| "num_layers": 2, |
| "hidden_act": "gelu", |
| "mm_tokens_per_image": 100, |
| "proj_dim": 1536, |
| "downsample_factor": 1 |
| }, |
| "lm_cfg": { |
| "model_type": "gemma4_text", |
| "data_type": "bfloat16", |
| "arch": "gemma", |
| "token_cfg": { |
| "vocab_size": 262144 |
| }, |
| "rope_cfg": { |
| "rope_theta": 1000000.0, |
| "rope_local_base_freq": 10000.0, |
| "rope_dimension_count": 128, |
| "sliding_rope_dimension_count": 256, |
| "rope_scaling": { |
| "factor": 1.0, |
| "low_freq_factor": 0, |
| "high_freq_factor": 0, |
| "original_max_position_embeddings": 0, |
| "long_factor": null, |
| "short_factor": null, |
| "rope_type": "proportional", |
| "mrope_section": null, |
| "mrope_interleaved": false |
| } |
| }, |
| "attn_cfg": { |
| "num_attention_heads": 8, |
| "num_key_value_heads": 1, |
| "head_dim": 512, |
| "sliding_head_dim": 256, |
| "swa_enable": true, |
| "sliding_window": 512, |
| "attention_bias": false, |
| "attention_dropout": 0.0, |
| "query_pre_attn_scalar": 0 |
| }, |
| "mlp_cfg": { |
| "intermediate_size": 6144, |
| "act": "gelu_pytorch_tanh", |
| "num_layers": 3, |
| "mlp_bias": false |
| }, |
| "hidden_size": 1536, |
| "num_hidden_layers": 35, |
| "max_position_embeddings": 2048, |
| "rms_norm_eps": 1e-06, |
| "rms_norm_unit_offset": false, |
| "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" |
| ], |
| "hidden_size_per_layer_input": 256, |
| "num_kv_shared_layers": 20, |
| "use_double_wide_mlp": true, |
| "attn_logit_softcapping": null, |
| "final_logit_softcapping": 30.0, |
| "lm_head_num_splits": 4, |
| "lm_head_split_dim": 65536, |
| "conv_L_cache": 3, |
| "conv_bias": false, |
| "lora_cfg": null, |
| "draft_cfg": null |
| }, |
| "pipeline_cfg": { |
| "system_prompt": null, |
| "chat_template": null, |
| "max_num_tokens": 2048, |
| "input_token_group_size": 128, |
| "input_token_group_offsets": [ |
| 0, |
| 128, |
| 256, |
| 384, |
| 512, |
| 640, |
| 768, |
| 896, |
| 1024, |
| 1152, |
| 1280, |
| 1408, |
| 1536, |
| 1664, |
| 1792, |
| 1920 |
| ], |
| "future_token_mask_size": 128, |
| "return_logits": false, |
| "use_strided_kv_cache": false, |
| "enable_filter_sharing": false, |
| "quantize_embeddings": false, |
| "split_mlp": true |
| }, |
| "language_model_name": "models--google--gemma-4-E2B-it_language", |
| "vision_model_name": "models--google--gemma-4-E2B-it_vision" |
| } |
|
|
