{ "architectures": [ "DiaForConditionalGeneration" ], "bos_token_id": null, "decoder_config": { "bos_token_id": 99, "cross_head_dim": 8, "cross_hidden_size": 16, "cross_num_attention_heads": 2, "cross_num_key_value_heads": 1, "eos_token_id": 97, "head_dim": 8, "hidden_act": "silu", "hidden_size": 32, "initializer_range": 0.02, "intermediate_size": 37, "max_position_embeddings": 50, "model_type": "dia_decoder", "norm_eps": 1e-05, "num_attention_heads": 2, "num_channels": 3, "num_hidden_layers": 2, "num_key_value_heads": 1, "pad_token_id": 98, "rope_parameters": { "rope_theta": 10000.0, "rope_type": "default" }, "use_cache": true, "vocab_size": 100 }, "delay_pattern": [ 0, 1, 2 ], "dtype": "float32", "encoder_config": { "head_dim": 8, "hidden_act": "silu", "hidden_size": 16, "initializer_range": 0.02, "intermediate_size": 37, "max_position_embeddings": 50, "model_type": "dia_encoder", "norm_eps": 1e-05, "num_attention_heads": 2, "num_hidden_layers": 2, "num_key_value_heads": 2, "rope_parameters": { "rope_theta": 10000.0, "rope_type": "default" }, "vocab_size": 100 }, "eos_token_id": null, "initializer_range": 0.02, "is_encoder_decoder": true, "model_type": "dia", "norm_eps": 1e-05, "pad_token_id": null, "transformers_version": "5.16.0.dev0", "use_cache": true }