idZddlmZddlmZddlmZddlmZm Z e je Z edeGd d eZ edeGd d eZedeGd deZgdZy)zDia model configuration)strict)PreTrainedConfig)RopeParameters)auto_docstringloggingznari-labs/Dia-1.6B) checkpointceZdZUdZdZeed<dZeed<dZeed<dZ eed<dZ eed <d Z eed <d Z eed <dZ eed<dZeed<dZeed<dZedzed<dZeed<y)DiaEncoderConfig dia_encodermax_position_embeddings num_hidden_layers hidden_sizenum_attention_headsnum_key_value_headshead_dimiintermediate_sizeh㈵>norm_eps vocab_sizesilu hidden_actNrope_parameters{Gz?initializer_range)__name__ __module__ __qualname__ model_typerint__annotations__rrrrrrrfloatrrstrrdictr z/var/www/vps2.regionflexible.com/Desarrollo/venv/lib/python3.12/site-packages/transformers/models/dia/configuration_dia.pyr r sJ#'S'sK!!!!Hc!s!HeJJ#'OTD['#u#r+r cteZdZUdZdZdZeed<dZeed<dZ eed<d Z eed <d Z eed <d Z eed<dZ eed<d Zeed<dZeed<d Zeed<dZeed<dZeed<dZeed<dZeed<dZeed<dZeezdzed<d Zeed!<d"Zeed#<d"Zeed$<d%Zedzed&<dZ ee!ezdzed'<d(Z"edzed)<y)*DiaDecoderConfiga) cross_num_attention_heads (`int`, *optional*, defaults to 16): Number of attention heads for each cross-attention layer in the Transformer decoder. cross_head_dim (`int`, *optional*, defaults to 128): Dimensionality of the cross-attention head. cross_num_key_value_heads (`int`, *optional*, defaults to 16): Number of key and value heads for each cross-attention layer in the Transformer decoder. cross_hidden_size (`int`, *optional*, defaults to 1024): Dimensionality of the cross-attention layers. dia_decoderi rriri rrrrrrcross_num_attention_headscross_head_dimcross_num_key_value_headsr cross_hidden_sizerrirrr num_channelsNrrr T use_cacheis_encoder_decoderi pad_token_id eos_token_idi bos_token_id)#r!r"r#__doc__r$rr%r&rrrrrrr2r3r4r5rr'rrr(r7rrr)r r8boolr9r:r;listr<r*r+r,r.r.-s J#'S'sK!s!!!  Hc%'s'NC%'s'!s!HeJJL#48O^d*T18#u#It###L#*#+/L#S /D(/#L#*#r+r.ceZdZUdZdZdgZeedZdZ ee zdze d<dZ ee zdze d<dZ ee d <d Zee d <dZedze d <dZeeezdze d <dZedze d<dZeedze d<dZee d<d Zee d<fdZdZdZxZS) DiaConfigaP delay_pattern (`list[int]`, *optional*, defaults to `[0, 8, 9, 10, 11, 12, 13, 14, 15]`): The delay pattern for the decoder. The length of this list must match `decoder_config.num_channels`. Example: ```python >>> from transformers import DiaConfig, DiaModel >>> # Initializing a DiaConfig with default values >>> configuration = DiaConfig() >>> # Initializing a DiaModel (with random weights) from the configuration >>> model = DiaModel(configuration) >>> # Accessing the model configuration >>> configuration = model.config ``` diapast_key_values)encoder_configdecoder_configNrDrErrTr9r:r;r< delay_patternrr r8c :t|jtrtdi|j|_t|jtrt di|j|_|j |jn t|_|j |jn t |_|j |j ngd|_|j0tjd|j|j_|j0tjd|j|j_ |j0tjd|j|j_ t|4di|y)N) rr6 r zjPassing `pad_token_id` to `DiaConfig` is deprecated. Please set it directly on `DiaDecoderConfig` instead.zjPassing `eos_token_id` to `DiaConfig` is deprecated. Please set it directly on `DiaDecoderConfig` instead.zjPassing `bos_token_id` to `DiaConfig` is deprecated. Please set it directly on `DiaDecoderConfig` instead.r*) isinstancerDr)r rEr.rFr:logger warning_oncer;r<super __post_init__)selfkwargs __class__s r,rRzDiaConfig.__post_init__{s` d))4 0"2"IT5H5H"ID  d))4 0"2"IT5H5H"ID 595H5H5Td11ZjZl595H5H5Td11ZjZl"&"4"4"@D  Fg     (   H 04/@/@D   ,    (   H 04/@/@D   ,    (   H 04/@/@D   , ''r+cr|jjt|jk7r t dy)zOPart of `@strict`-powered validation. Validates the architecture of the config.z3Number of channels must match delay pattern length.N)rEr7lenrF ValueError)rSs r,validate_architecturezDiaConfig.validate_architectures3    + +s43E3E/F FRS S Gr+c|jS)z^Defaulting to audio config as it's the decoder in this case which is usually the text backbone)rE)rSargsrTs r,get_text_configzDiaConfig.get_text_configs"""r+)r!r"r#r=r$keys_to_ignore_at_inferencer r. sub_configsrDr)r&rErr'r9r>r:r%r;r?r<rFr r8rRrYr\ __classcell__)rUs@r,rArAUs(J#4"5%5IYZK59N$t+d2959N$t+d29He###L#*#+/L#S /D(/#L#*#&*M49t#*#u#It"(HT #r+rA)rAr r.N)r=huggingface_hub.dataclassesrconfiguration_utilsrmodeling_rope_utilsrutilsrr get_loggerr!rOr r.rA__all__r*r+r,rfs.31,   H %/0$'$1$"/0#$'#$1#$L/0O# O#1O#d @r+