mjf-su/PhysicalAI-base-VLA
016
1{2 "architectures": [3 "Qwen3VLForConditionalGeneration"4 ],5 "dtype": "bfloat16",6 "eos_token_id": 151645,7 "image_token_id": 151655,8 "model_type": "qwen3_vl",9 "pad_token_id": 151643,10 "text_config": {11 "attention_bias": false,12 "attention_dropout": 0.0,13 "bos_token_id": 151643,14 "dtype": "bfloat16",15 "eos_token_id": 151645,16 "head_dim": 128,17 "hidden_act": "silu",18 "hidden_size": 2560,19 "initializer_range": 0.02,20 "intermediate_size": 9728,21 "max_position_embeddings": 262144,22 "model_type": "qwen3_vl_text",23 "num_attention_heads": 32,24 "num_hidden_layers": 36,25 "num_key_value_heads": 8,26 "rms_norm_eps": 1e-06,27 "rope_scaling": {28 "mrope_interleaved": true,29 "mrope_section": [30 24,31 20,32 2033 ],34 "rope_type": "default"35 },36 "rope_theta": 5000000,37 "tie_word_embeddings": true,38 "use_cache": true,39 "vocab_size": 15193640 },41 "tie_word_embeddings": true,42 "transformers_version": "4.57.1",43 "use_cache": false,44 "video_token_id": 151656,45 "vision_config": {46 "deepstack_visual_indexes": [47 5,48 11,49 1750 ],51 "depth": 24,52 "dtype": "bfloat16",53 "hidden_act": "gelu_pytorch_tanh",54 "hidden_size": 1024,55 "in_channels": 3,56 "initializer_range": 0.02,57 "intermediate_size": 4096,58 "model_type": "qwen3_vl",59 "num_heads": 16,60 "num_position_embeddings": 2304,61 "out_hidden_size": 2560,62 "patch_size": 16,63 "spatial_merge_size": 2,64 "temporal_patch_size": 265 },66 "vision_end_token_id": 151653,67 "vision_start_token_id": 15165268}69 