mlx-community/MolmoPoint-8B-5bit

1{2    "adapter_config": {3        "attention_dropout": 0.0,4        "attn_implementation": "sdpa",5        "float32_attention": true,6        "head_dim": 72,7        "hidden_act": "silu",8        "hidden_size": 1152,9        "image_feature_dropout": 0,10        "initializer_range": 0.02,11        "intermediate_size": 12288,12        "model_type": "molmo_point",13        "num_attention_heads": 16,14        "num_key_value_heads": 16,15        "pooling_attention_mask": true,16        "positional_embeddings": null,17        "residual_dropout": 0.0,18        "text_hidden_size": 4096,19        "vit_layers": [20            -3,21            -922        ]23    },24    "architectures": [25        "MolmoPointForConditionalGeneration"26    ],27    "auto_map": {28        "AutoConfig": "configuration_molmo_point.MolmoPointConfig",29        "AutoModelForImageTextToText": "modeling_molmo_point.MolmoPointForConditionalGeneration"30    },31    "dtype": "float32",32    "embed_location": false,33    "embed_selected_vit_patch": "linear",34    "eos_token_id": 151645,35    "frame_end_token_id": 151944,36    "frame_start_token_id": 151943,37    "image_col_id": 151939,38    "image_end_token_id": 151937,39    "image_high_res_id": 151938,40    "image_non_indexable_patch_id": 151942,41    "image_patch_id": 151938,42    "image_start_token_id": 151936,43    "initializer_range": 0.02,44    "layer_norm_x": true,45    "location_token_id": 151949,46    "low_res_image_start_token_id": null,47    "mask_patches": "always",48    "mask_repeats": "inference",49    "mask_subpatches": "inference",50    "model_type": "molmo_point",51    "no_more_points_class": true,52    "norm_logits": true,53    "patch_embed_dim": 512,54    "patch_embedding_kind": "image_feature0",55    "patch_location": "3x3",56    "patch_token_id": 151947,57    "quantization": {58        "group_size": 64,59        "bits": 5,60        "mode": "affine"61    },62    "quantization_config": {63        "group_size": 64,64        "bits": 5,65        "mode": "affine"66    },67    "subpatch_token_id": 151948,68    "text_config": {69        "additional_vocab_size": 128,70        "attention_dropout": 0.0,71        "attn_implementation": "sdpa",72        "embedding_dropout": 0.0,73        "head_dim": 128,74        "hidden_act": "silu",75        "hidden_size": 4096,76        "initializer_range": 0.02,77        "intermediate_size": 12288,78        "layer_norm_eps": 1e-06,79        "max_position_embeddings": 37376,80        "model_type": "molmo2_text",81        "norm_after": false,82        "num_attention_heads": 32,83        "num_hidden_layers": 36,84        "num_key_value_heads": 8,85        "qk_norm_type": "qwen3",86        "qkv_bias": false,87        "residual_dropout": 0.0,88        "rope_scaling": null,89        "rope_scaling_layers": null,90        "rope_theta": 1000000.0,91        "use_cache": true,92        "use_qk_norm": true,93        "vocab_size": 15193694    },95    "tie_word_embeddings": false,96    "token_prediction_rotary": "one_d",97    "token_prediction_rotary_theta": 50000.0,98    "transformers_version": "4.57.1",99    "use_cache": true,100    "use_frame_special_tokens": true,101    "vision_config": {},102    "vit_config": {103        "attention_dropout": 0.0,104        "attn_implementation": "sdpa",105        "float32_attention": true,106        "head_dim": 72,107        "hidden_act": "gelu_pytorch_tanh",108        "hidden_size": 1152,109        "image_default_input_size": [110            378,111            378112        ],113        "image_num_pos": 729,114        "image_patch_size": 14,115        "initializer_range": 0.02,116        "intermediate_size": 4304,117        "layer_norm_eps": 1e-06,118        "model_type": "molmo2",119        "num_attention_heads": 16,120        "num_hidden_layers": 27,121        "num_key_value_heads": 16,122        "residual_dropout": 0.0123    }124}