prithivMLmods/proxima-ocr-d.markdown-post3.0.l
519
1{2 "architectures": [3 "Qwen3VLForConditionalGeneration"4 ],5 "dtype": "bfloat16",6 "image_token_id": 151655,7 "max_seq_length": 2048,8 "model_type": "qwen3_vl",9 "text_config": {10 "attention_bias": false,11 "attention_dropout": 0.0,12 "bos_token_id": 151643,13 "dtype": "bfloat16",14 "eos_token_id": 151645,15 "head_dim": 128,16 "hidden_act": "silu",17 "hidden_size": 4096,18 "initializer_range": 0.02,19 "intermediate_size": 12288,20 "max_position_embeddings": 262144,21 "model_type": "qwen3_vl_text",22 "num_attention_heads": 32,23 "num_hidden_layers": 36,24 "num_key_value_heads": 8,25 "rms_norm_eps": 1e-06,26 "rope_scaling": {27 "mrope_interleaved": true,28 "mrope_section": [29 24,30 20,31 2032 ],33 "rope_type": "default"34 },35 "rope_theta": 5000000,36 "use_cache": true,37 "vocab_size": 15193638 },39 "tie_word_embeddings": false,40 "transformers_version": "4.57.1",41 "use_cache": false,42 "use_cut_cross_entropy": false,43 "video_token_id": 151656,44 "vision_config": {45 "deepstack_visual_indexes": [46 8,47 16,48 2449 ],50 "depth": 27,51 "dtype": "bfloat16",52 "hidden_act": "gelu_pytorch_tanh",53 "hidden_size": 1152,54 "in_channels": 3,55 "initializer_range": 0.02,56 "intermediate_size": 4304,57 "model_type": "qwen3_vl",58 "num_heads": 16,59 "num_position_embeddings": 2304,60 "out_hidden_size": 4096,61 "patch_size": 16,62 "spatial_merge_size": 2,63 "temporal_patch_size": 264 },65 "vision_end_token_id": 151653,66 "vision_start_token_id": 15165267}68 