| { | |
| "architectures": [ | |
| "XOmniForCausalLM" | |
| ], | |
| "attention_dropout": 0.0, | |
| "auto_map": { | |
| "AutoConfig": "modeling_xomni.XOmniConfig", | |
| "AutoModel": "modeling_xomni.XOmniModel", | |
| "AutoModelForCausalLM": "modeling_xomni.XOmniForCausalLM" | |
| }, | |
| "eos_token_id": 151643, | |
| "hidden_act": "silu", | |
| "hidden_size": 3584, | |
| "image_vocab_size": 16384, | |
| "initializer_range": 0.02, | |
| "intermediate_size": 18944, | |
| "max_position_embeddings": 8192, | |
| "max_window_layers": 28, | |
| "mm_special_tokens": [ | |
| "<SOM>", | |
| "<EOM>", | |
| "<IMAGE>" | |
| ], | |
| "mm_vocab_size": 16448, | |
| "model_type": "x-omni", | |
| "num_attention_heads": 28, | |
| "num_hidden_layers": 36, | |
| "num_key_value_heads": 4, | |
| "num_mm_adap_layers": 4, | |
| "num_mm_head_layers": 4, | |
| "pad_token_id": 151643, | |
| "rms_norm_eps": 1e-06, | |
| "rope_scaling": null, | |
| "rope_theta": 1000000.0, | |
| "sliding_window": null, | |
| "tie_word_embeddings": false, | |
| "torch_dtype": "bfloat16", | |
| "transformers_version": "4.52.0", | |
| "use_cache": true, | |
| "use_sliding_window": false, | |
| "vision_config":{ | |
| "transform": { | |
| "short_size": 384, | |
| "long_size": 1152, | |
| "patch_size": 16, | |
| "random_ratio": null, | |
| "min_short_size": 128, | |
| "max_aspect_ratio": 3.0, | |
| "filtering": false | |
| }, | |
| "encoder": { | |
| "siglip_name": "siglip2_giant_patch16_384", | |
| "siglip_path": "vit/vit_g.pth", | |
| "projector_path": "vit/siglip_vq.pt", | |
| "with_norm": true, | |
| "z_channels": 1536, | |
| "codebook_size": 16384, | |
| "codebook_dim": 2048 | |
| }, | |
| "decoder": { | |
| "model_path": "diffusers", | |
| "num_inference_steps": 28, | |
| "cfg_scale": 1.5, | |
| "cfg_scale_2": 1.5, | |
| "upscale_factor": 16 | |
| }, | |
| "dtype": "bfloat16" | |
| }, | |
| "vocab_size": 151936 | |
| } | |