qiO6dZddlZddlZddlmZddlmZddlmZddl Z ddl m cm Z ddl m Z ddlmZdd lmZdd lmZdd lmZmZmZdd lmZmZdd lmZddlmZddl m!Z!m"Z"m#Z#m$Z$m%Z%m&Z&ddl'm(Z(m)Z)m*Z*e%jVe,Z-dZ.dZ/dZ0de jbde jbfdZ2ee#dGdde!Z3ee#dGdde!Z4ee#Gdd e!Z5Gd!d"e jlZ7Gd#d$e jlZ8Gd%d&e jlZ9Gd'd(e jlZ:Gd)d*e jlZ;Gd+d,e jlZ<Gd-d.e jlZ=Gd/d0e jlZ>Gd1d2e jlZ?Gd3d4eZ@Gd5d6e jlZAGd7d8e jlZBGd9d:e jlZCGd;de jbd?e jbd@e jbdAe jbdzdBeEdCeEfdDZFGdEdFe jlZGGdGdHe jlZHGdIdJe jlZIGdKdLe jlZJGdMdNe jlZKGdOdPeZLGdQdRe jlZMGdSdTe jlZNe#GdUdVeZOGdWdXeOZPe#dYGdZd[eOZQe#Gd\d]eOZRe#Gd^d_eOZSe#Gd`daeOZTgdbZUy)dzPyTorch CLAP model.N)Callable) dataclass)Any)nn)initialization)ACT2FN)GradientCheckpointingLayer)BaseModelOutputBaseModelOutputWithPooling,BaseModelOutputWithPoolingAndCrossAttentions)ALL_ATTENTION_FUNCTIONSPreTrainedModel)Unpack)apply_chunking_to_forward) 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 X/opt/pipecat/venv/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_partitionr7>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.)r6r2r3r4r5s r&window_reverser9Ssn==$Lll2v4e{6JKYdfrsGooaAq!Q/::<AA"feUabG Nr(logitsreturnctjt||j}tj j ||S)Ndevice)torcharangelenr>r functional cross_entropy)r:labelss r&contrastive_lossrEhs1 \\#f+fmm