i6dZddlZddlZddlmZddlmZddlmZddl Z ddl m cm Z ddl m Z ddlmZdd lmZdd lmZdd lmZmZmZdd lmZmZdd lmZddlmZm Z ddl!m"Z"m#Z#m$Z$m%Z%m&Z&m'Z'ddl(m)Z)m*Z*m+Z+e&jXe-Z.dZ/dZ0dZ1de jdde jdfdZ3ee$dGdde"Z4ee$dGdde"Z5ee$Gdd e"Z6Gd!d"e jnZ8Gd#d$e jnZ9Gd%d&e jnZ:Gd'd(e jnZ;Gd)d*e jnZ<Gd+d,e jnZ=Gd-d.e jnZ>Gd/d0e jnZ?Gd1d2e jnZ@Gd3d4eZAGd5d6e jnZBGd7d8e jnZCGd9d:e jnZDGd;de jdd?e jdd@e jddAe jddzdBeFdCeFfdDZGGdEdFe jnZHGdGdHe jnZIGdIdJe jnZJGdKdLe jnZKGdMdNe jnZLGdOdPeZMGdQdRe jnZNGdSdTe jnZOe$GdUdVeZPGdWdXePZQe$dYGdZd[ePZRe$Gd\d]ePZSe$Gd^d_ePZTe$Gd`daePZUgdbZVy)dzPyTorch CLAP model.N)Callable) dataclass)Any)nn)initialization)ACT2FN)GradientCheckpointingLayer)BaseModelOutputBaseModelOutputWithPooling,BaseModelOutputWithPoolingAndCrossAttentions)ALL_ATTENTION_FUNCTIONSPreTrainedModel)Unpack)apply_chunking_to_forwardmeshgrid) ModelOutputTransformersKwargsauto_docstringcan_return_tuplelogging torch_int)ClapAudioConfig ClapConfigClapTextConfigc|j\}}}|dddddddfjdd|d}|j|||z|}|S)ae Interpolate data in time domain. This is used to compensate the resolution reduction in downsampling of a CNN. Args: hidden_states (`torch.FloatTensor` of shape (batch_size, time_length, classes_num)): Input hidden states ratio (`int`): The ratio of the length of the output to the length of the input. Nr)shaperepeatreshape) hidden_statesratio batch_size time_length classes_num upsampleds p/home/obispo/Crisostomo_bridge/mision_env/lib/python3.12/site-packages/transformers/models/clap/modeling_clap.py interpolater(-sX.;-@-@*ZkaD!m,33Aq%CI!!*kE.A;OI c|j\}}}}|j|||z|||z||}|jddddddjjd|||}|S)aR Returns the resized hidden states. The output shape should be `(batch_size * num_windows, window_size, window_size, num_channels)` Args: hidden_states (`torch.FloatTensor` of shape `(batch_size, height, width, num_channels)`): Input hidden states window_size (`int`): Window size rrrrviewpermute contiguous)r! window_sizer#heightwidth num_channelswindowss r'window_partitionr8>s}/<.A.A+J|!&&Fk);8Lk[gM##Aq!Q15@@BGGKYdfrsG Nr)c|jd}|jd||z||z|||}|jddddddjjd|||}|S)a Merges windows to produce higher resolution features. Args: windows (`torch.FloatTensor` of shape `(num_windows * batch_size, window_size, window_size, num_channels)`): Input windows window_size (`int`): Window size height (`int`): Height of the resized audio width (`int`): Width of the resized audio r.rrrr+r,r-r/)r7r3r4r5r6s r'window_reverser:Ssn==$Lll2v4e{6JKYdfrsGooaAq!Q/::<AA"feUabG Nr)logitsreturnctjt||j}tj j ||S)Ndevice)torcharangelenr?r functional cross_entropy)r;labelss r'contrastive_lossrFhs1 \\#f+fmm