{ | |
"attn_dropout_p": 0.0, | |
"class_dropout_prob": 0.1, | |
"cls_token_num": 1, | |
"dim": 1280, | |
"drop_path_rate": 0.0, | |
"ffn_dim_multiplier": null, | |
"ffn_dropout_p": 0.1, | |
"frame_prediction": false, | |
"initializer_range": 0.02, | |
"max_batch_size": 32, | |
"max_seq_len": 1024, | |
"model_type": "class_cond", | |
"multiple_of": 256, | |
"n_head": 20, | |
"n_kv_head": null, | |
"n_layer": 30, | |
"norm_eps": 1e-05, | |
"num_classes": 101, | |
"resid_dropout_p": 0.1, | |
"rope_base": 10000, | |
"token_dropout_p": 0.1, | |
"use_fixed_pe": false, | |
"vocab_size": 8192 | |
} |