text_encoder/config.json
| 1 | { |
| 2 | "architectures": [ |
| 3 | "Qwen3VLForConditionalGeneration" |
| 4 | ], |
| 5 | "dtype": "bfloat16", |
| 6 | "image_token_id": 151655, |
| 7 | "model_type": "qwen3_vl", |
| 8 | "text_config": { |
| 9 | "attention_bias": false, |
| 10 | "attention_dropout": 0.0, |
| 11 | "bos_token_id": 151643, |
| 12 | "dtype": "bfloat16", |
| 13 | "eos_token_id": 151645, |
| 14 | "head_dim": 128, |
| 15 | "hidden_act": "silu", |
| 16 | "hidden_size": 4096, |
| 17 | "initializer_range": 0.02, |
| 18 | "intermediate_size": 12288, |
| 19 | "max_position_embeddings": 262144, |
| 20 | "model_type": "qwen3_vl_text", |
| 21 | "num_attention_heads": 32, |
| 22 | "num_hidden_layers": 36, |
| 23 | "num_key_value_heads": 8, |
| 24 | "rms_norm_eps": 1e-06, |
| 25 | "rope_scaling": { |
| 26 | "mrope_interleaved": true, |
| 27 | "mrope_section": [ |
| 28 | 24, |
| 29 | 20, |
| 30 | 20 |
| 31 | ], |
| 32 | "rope_type": "default" |
| 33 | }, |
| 34 | "rope_theta": 5000000, |
| 35 | "use_cache": true, |
| 36 | "vocab_size": 151936 |
| 37 | }, |
| 38 | "tie_word_embeddings": false, |
| 39 | "transformers_version": "4.57.1", |
| 40 | "video_token_id": 151656, |
| 41 | "vision_config": { |
| 42 | "deepstack_visual_indexes": [ |
| 43 | 8, |
| 44 | 16, |
| 45 | 24 |
| 46 | ], |
| 47 | "depth": 27, |
| 48 | "dtype": "bfloat16", |
| 49 | "hidden_act": "gelu_pytorch_tanh", |
| 50 | "hidden_size": 1152, |
| 51 | "in_channels": 3, |
| 52 | "initializer_range": 0.02, |
| 53 | "intermediate_size": 4304, |
| 54 | "model_type": "qwen3_vl", |
| 55 | "num_heads": 16, |
| 56 | "num_position_embeddings": 2304, |
| 57 | "out_hidden_size": 4096, |
| 58 | "patch_size": 16, |
| 59 | "spatial_merge_size": 2, |
| 60 | "temporal_patch_size": 2 |
| 61 | }, |
| 62 | "vision_end_token_id": 151653, |
| 63 | "vision_start_token_id": 151652 |
| 64 | } |
| 65 | |