{ | |
"_name_or_path": "Audiogen/apolloxl-pretrain-030724", | |
"ada_dim": 512, | |
"architectures": [ | |
"Apollo" | |
], | |
"codec_config": "625e102f2cf534109a62d19b484d5259222fcaa7", | |
"codec_mean": -0.09, | |
"codec_name": "vae-100hz-32ch-030624", | |
"codec_revision": "625e102f2cf534109a62d19b484d5259222fcaa7", | |
"codec_std": 1.215, | |
"conditioning_config": { | |
"cross_cond_dims": {}, | |
"cross_sequence_length": 64, | |
"embed_cond_dims": {}, | |
"global_cond_dims": { | |
"pooled_audio": 512 | |
}, | |
"multimodal_layer_factor": 4, | |
"multimodal_query_sharing": true | |
}, | |
"eps": 1e-08, | |
"head_dim": 128, | |
"hidden_dim": 3072, | |
"input_dim": 32, | |
"intermediate_ada_dim": 1280, | |
"intermediate_dim": 8192, | |
"model_type": "apollo", | |
"num_heads": 24, | |
"num_layers": 24, | |
"patch_size": 2, | |
"rope_base": 1000000, | |
"rope_n_elem": 128, | |
"sequence_length": 3072, | |
"time_dim": 512, | |
"torch_dtype": "bfloat16", | |
"transformers_version": "4.43.0.dev0" | |
} | |