| { | |
| "architectures": [ | |
| "OpenPanguVLForConditionalGeneration" | |
| ], | |
| "bias": true, | |
| "auto_map": { | |
| "AutoConfig": "configuration_openpangu_vl.OpenPanguVLConfig", | |
| "AutoModel": "modeling_openpangu_vl.OpenPanguVLModel", | |
| "AutoModelForCausalLM": "modeling_openpangu_vl.OpenPanguVL" | |
| }, | |
| "attention_dropout": 0.0, | |
| "bos_token_id": 1, | |
| "eos_token_id": 45892, | |
| "vision_start_token_id": 45986, | |
| "vision_end_token_id": 46007, | |
| "image_token_id": 46005, | |
| "video_token_id": 144208, | |
| "hidden_act": "silu", | |
| "hidden_size": 4096, | |
| "initializer_range": 0.02, | |
| "intermediate_size": 12800, | |
| "max_position_embeddings": 131072, | |
| "max_window_layers": 34, | |
| "model_type": "openpangu_vl", | |
| "num_attention_heads": 32, | |
| "num_hidden_layers": 34, | |
| "num_key_value_heads": 8, | |
| "rms_norm_eps": 1e-05, | |
| "rope_theta": 64000000.0, | |
| "tie_word_embeddings": false, | |
| "torch_dtype": "bfloat16", | |
| "transformers_version": "4.52.4", | |
| "use_cache": true, | |
| "use_sliding_window": false, | |
| "vision_config": { | |
| "rms_norm_eps": 1e-06, | |
| "depth": 26, | |
| "hidden_act": "gelu", | |
| "hidden_size": 1536, | |
| "intermediate_size": 4608, | |
| "num_heads": 16, | |
| "in_chans": 3, | |
| "out_hidden_size": 3584, | |
| "patch_size": 14, | |
| "spatial_merge_size": 2, | |
| "spatial_patch_size": 14, | |
| "window_size": 112, | |
| "fullatt_block_indexes": [ | |
| 5, | |
| 12, | |
| 19, | |
| 25 | |
| ], | |
| "tokens_per_second": 2, | |
| "temporal_patch_size": 2 | |
| }, | |
| "rope_scaling": { | |
| "rope_type": "default", | |
| "type": "default", | |
| "mrope_section": [ | |
| 10, | |
| 27, | |
| 27 | |
| ], | |
| "mrope_interleaved": true | |
| }, | |
| "vocab_size": 153376 | |
| } |