ByteDance/EvoQuality
30200
1{2 "architectures": [3 "Qwen2_5_VLForConditionalGeneration"4 ],5 "attention_dropout": 0.0,6 "eos_token_id": 151645,7 "hidden_act": "silu",8 "hidden_size": 3584,9 "image_token_id": 151655,10 "initializer_range": 0.02,11 "intermediate_size": 18944,12 "max_position_embeddings": 128000,13 "max_window_layers": 28,14 "model_type": "qwen2_5_vl",15 "num_attention_heads": 28,16 "num_hidden_layers": 28,17 "num_key_value_heads": 4,18 "pad_token_id": 151643,19 "rms_norm_eps": 1e-06,20 "rope_scaling": {21 "mrope_section": [22 16,23 24,24 2425 ],26 "rope_type": "default",27 "type": "default"28 },29 "rope_theta": 1000000.0,30 "sliding_window": 32768,31 "tie_word_embeddings": false,32 "torch_dtype": "float32",33 "transformers_version": "4.51.3",34 "use_cache": true,35 "use_sliding_window": false,36 "video_token_id": 151656,37 "vision_config": {38 "depth": 32,39 "fullatt_block_indexes": [40 7,41 15,42 23,43 3144 ],45 "hidden_act": "silu",46 "hidden_size": 1280,47 "in_channels": 3,48 "in_chans": 3,49 "intermediate_size": 3420,50 "model_type": "qwen2_5_vl",51 "num_heads": 16,52 "out_hidden_size": 3584,53 "patch_size": 14,54 "spatial_merge_size": 2,55 "spatial_patch_size": 14,56 "temporal_patch_size": 2,57 "tokens_per_second": 2,58 "torch_dtype": "float32",59 "window_size": 11260 },61 "vision_end_token_id": 151653,62 "vision_start_token_id": 151652,63 "vision_token_id": 151654,64 "vocab_size": 15206465}66 