prithivMLmods/oMEGA-4B-SpatialThink-0804

1{2  "architectures": [3    "Qwen3VLForConditionalGeneration"4  ],5  "bos_token_id": null,6  "dtype": "bfloat16",7  "eos_token_id": 151645,8  "image_token_id": 151655,9  "model_type": "qwen3_vl",10  "pad_token_id": 151643,11  "text_config": {12    "attention_bias": false,13    "attention_dropout": 0.0,14    "bos_token_id": 151643,15    "dtype": "bfloat16",16    "eos_token_id": 151645,17    "head_dim": 128,18    "hidden_act": "silu",19    "hidden_size": 2560,20    "initializer_range": 0.02,21    "intermediate_size": 9728,22    "max_position_embeddings": 262144,23    "model_type": "qwen3_vl_text",24    "num_attention_heads": 32,25    "num_hidden_layers": 36,26    "num_key_value_heads": 8,27    "pad_token_id": null,28    "rms_norm_eps": 1e-06,29    "rope_parameters": {30      "mrope_interleaved": true,31      "mrope_section": [32        24,33        20,34        2035      ],36      "rope_theta": 5000000,37      "rope_type": "default"38    },39    "tie_word_embeddings": true,40    "use_cache": true,41    "vocab_size": 15193642  },43  "tie_word_embeddings": true,44  "transformers_version": "5.14.1",45  "video_token_id": 151656,46  "vision_config": {47    "deepstack_visual_indexes": [48      5,49      11,50      1751    ],52    "depth": 24,53    "dtype": "bfloat16",54    "hidden_act": "gelu_pytorch_tanh",55    "hidden_size": 1024,56    "in_channels": 3,57    "initializer_range": 0.02,58    "intermediate_size": 4096,59    "model_type": "qwen3_vl_vision",60    "num_heads": 16,61    "num_position_embeddings": 2304,62    "out_hidden_size": 2560,63    "patch_size": 16,64    "spatial_merge_size": 2,65    "temporal_patch_size": 266  },67  "vision_end_token_id": 151653,68  "vision_start_token_id": 15165269}70