39 lines
959 B
JSON
39 lines
959 B
JSON
{
|
|
"return_dict": true,
|
|
"output_hidden_states": false,
|
|
"torchscript": false,
|
|
"dtype": "bfloat16",
|
|
"tie_word_embeddings": false,
|
|
"is_encoder_decoder": false,
|
|
"is_decoder": false,
|
|
"add_cross_attention": false,
|
|
"architectures": [
|
|
"YuE2ForCausalLM"
|
|
],
|
|
"bos_token_id": null,
|
|
"pad_token_id": null,
|
|
"eos_token_id": null,
|
|
"decoder_start_token_id": null,
|
|
"transformers_version": "4.57.6",
|
|
"auto_map": {
|
|
"AutoConfig": "modeling_yue2.YuE2Config",
|
|
"AutoModelForCausalLM": "modeling_yue2.YuE2ForCausalLM"
|
|
},
|
|
"model_type": "yue2",
|
|
"hidden_size": 2048,
|
|
"num_hidden_layers": 28,
|
|
"num_attention_heads": 16,
|
|
"num_key_value_heads": 8,
|
|
"head_dim": 128,
|
|
"intermediate_size": 6144,
|
|
"vocab_size": 184704,
|
|
"rms_norm_eps": 1e-06,
|
|
"rope_theta": 1000000,
|
|
"max_position_embeddings": 24576,
|
|
"latent_type": "vae",
|
|
"latent_dim": 64,
|
|
"max_latent_frames": 24576,
|
|
"timestep_shift": 1.0,
|
|
"output_attentions": false
|
|
}
|