| { | |
| "architectures": [ | |
| "MiMoVisionModel" | |
| ], | |
| "auto_map": { | |
| "AutoConfig": "modeling_mimo_vision.MiMoVisionConfig", | |
| "AutoModel": "modeling_mimo_vision.MiMoVisionModel" | |
| }, | |
| "depth": 28, | |
| "dtype": "bfloat16", | |
| "fullatt_block_indexes": [ | |
| 0, | |
| 9, | |
| 18, | |
| 27 | |
| ], | |
| "hidden_act": "silu", | |
| "hidden_size": 1280, | |
| "in_chans": 3, | |
| "intermediate_size": 4608, | |
| "model_type": "mimo26_vision", | |
| "num_heads": 32, | |
| "num_key_value_heads": 8, | |
| "num_query_groups": 4, | |
| "out_hidden_size": 6144, | |
| "patch_size": 16, | |
| "qk_channels": 64, | |
| "rms_norm_eps": 1e-06, | |
| "spatial_merge_size": 2, | |
| "spatial_patch_size": 16, | |
| "temporal_patch_size": 2, | |
| "tokens_per_second": 2, | |
| "transformers_version": "5.17.0", | |
| "use_sink": true, | |
| "visual_token_window_size": 64, | |
| "vit_window_attn_types": [ | |
| -1, | |
| 0, | |
| 0, | |
| 0, | |
| 0, | |
| 1, | |
| 1, | |
| 1, | |
| 1, | |
| -1, | |
| 0, | |
| 0, | |
| 0, | |
| 0, | |
| 1, | |
| 1, | |
| 1, | |
| 1, | |
| -1, | |
| 0, | |
| 0, | |
| 0, | |
| 0, | |
| 1, | |
| 1, | |
| 1, | |
| 1, | |
| -1 | |
| ], | |
| "window_size": 128 | |
| } | |