fTh^^SrSSKrSSKrSSKJr SSKJrJrJrJ r J r SSK r SSK r SSK J r SSKJr SSKJrJr SS KJrJr SS KJrJr SS KJrJrJrJr SS KJr SS K J!r!J"r" SSK#J$r$J%r%J&r& SSK'J(r(J)r)J*r* \&RV"\,5r-S\ R\S\ R\4Sjr/S\ R\S\ R\4Sjr0Sr1SASjr2SBSjr3\"SS\$55r4\"SS\$55r5"SS\ Rl5r7"SS \ Rl5r8"S!S"\ Rl5r9"S#S$\ Rl5r:"S%S&\ Rl5r;"S'S(\ Rl5r<"S)S*\ Rl5r="S+S,\ Rl5r>"S-S.\ Rl5r?"S/S0\ Rl5r@\%"S1S2\55rA"S3S4\A5rB"S5S6\A5rC\%"S7S8\A55rD\%"S9S:9"S;S<\A\55rE\%"S=S:9"S>S?\A\55rF/S@QrGg)CzPyTorch CLVP model.N) dataclass)CallableDictOptionalTupleUnion)nn)CrossEntropyLoss)ACT2FNget_activation)GenerationConfigGenerationMixin)_prepare_4d_attention_mask!_prepare_4d_causal_attention_mask)BaseModelOutput)BaseModelOutputWithPastAndCrossAttentionsBaseModelOutputWithPooling!CausalLMOutputWithCrossAttentions)PreTrainedModel)Conv1Disin_mps_friendly) ModelOutputauto_docstringlogging) ClvpConfigClvpDecoderConfigClvpEncoderConfiglogitsreturnc[RRU[R"[ U5UR S95$)Ndevice)r functional cross_entropytorcharangelenr$)r s ^/var/www/auris/envauris/lib/python3.13/site-packages/transformers/models/clvp/modeling_clvp.pycontrastive_lossr+7s/ == & &vu||CKPVP]P]/^ __ similaritycX[U5n[UR55nX-S- $)Ng@)r+t)r- caption_loss speech_losss r* clvp_lossr2<s*#J/L":<<>2K  &# --r,cUSSURSS-24nUSURSS-S24n[R"U*U4SS9$)z*Rotates half the hidden dims of the input..Ndim)shaper'cat)xx1x2s r* rotate_halfr=CsZ 3"!''"+"" " #B 3 q " " #B 99rc2YB ''r,cX5RU5nXERU5nX-[U5U--nX-[U5U--nX#-[U5U--n XxU 4$)a&Applies Rotary Position Embedding to the query and key tensors. Args: q (`torch.Tensor`): The query tensor. k (`torch.Tensor`): The key tensor. cos (`torch.Tensor`): The cosine part of the rotary embedding. sin (`torch.Tensor`): The sine part of the rotary embedding. position_ids (`torch.Tensor`): The position indices of the tokens corresponding to the query and key tensors. For example, this can be used to pass offsetted position ids when working with a KV-cache. unsqueeze_dim (`int`, *optional*, defaults to 1): The 'unsqueeze_dim' argument specifies the dimension along which to unsqueeze cos[position_ids] and sin[position_ids] so that they can be properly broadcasted to the dimensions of q and k. For example, note that cos[position_ids] and sin[position_ids] have the shape [batch_size, seq_len, head_dim]. Then, if q and k have the shape [batch_size, heads, seq_len, head_dim], then setting unsqueeze_dim=1 makes cos[position_ids] and sin[position_ids] broadcastable to the shapes of q and k. Similarly, if q and k have the shape [batch_size, seq_len, heads, head_dim], then set unsqueeze_dim=2. Returns: `tuple(torch.Tensor)` comprising of the query and key tensors rotated using the Rotary Position Embedding. ) unsqueezer=) qkvcossin position_ids unsqueeze_dimq_embedk_embedv_embeds r*apply_rotary_pos_embrJJst*   % %m 4C  % %m 4Cw;q>C/0Gw;q>C/0Gw;q>C/0G W $$r,c RU(aW[RRRUSUS9nUb)[RRRUSSS9OUnUnU(Ga=[R"UR SUR SS-4UR URS9n[U5Hup[X5R5(ai[R"X:H5SR5n [R"U SU [R"U/URS9XS/5Xx'M[RRRU SUS9Xx'M Ub)[RRRUSSS9OUnXq4$) z This method adds extra bos and eos tokens to input_ids and accordingly modifies the attention_mask which is used in `ClvpConditioningEncoder` and the generation loop of the `ClvpModelForConditionalGeneration`. )rrvalueNrrdtyper$r#)rr)r'r r%padzerosr8rOr$ enumeratersumwheremin concatenatetensor) input_idsattention_mask pad_token_id bos_token_id eos_token_id add_bos_token add_eos_tokenmodified_input_idsi each_input_idposs r*_pad_extra_bos_eos_tokensrcgsHH''++Iv\+R HVHbEHH   # #NF! # Dhv #"[[ __Q !3a!7 8 XaXhXh !*) 4 A =AACCkk-"?@CGGI(-(9(9"4C(%,, ~iN^N^*_anosatu)"% ).(;(;(?(? v]i(?(j"%!5IWHbEHH   # #NF! # Dhv   --r,c\rSrSr%SrSr\\R\ S'Sr \\R\ S'Sr \\R\ S'Sr \\ \R\ S'Sr\\ \R\ S'S rg) ClvpEncoderOutputa# Base class for CLVP encoder's outputs that contains a pooling of the last hidden states as well as a projection output (a linear layer on top of the pooled output). Args: embeds (`torch.FloatTensor` of shape `(batch_size, output_dim)`, *optional*, returned when model is initialized with `with_projection=True`): The embeddings obtained by applying the projection layer to the pooler_output. last_hidden_state (`torch.FloatTensor` of shape `(batch_size, sequence_length, hidden_size)`): The hidden state of the last layer of the model. pooler_output (`torch.FloatTensor` of shape `(batch_size, hidden_size)`): Pooled output of the `last_hidden_state`. hidden_states (`tuple(torch.FloatTensor)`, *optional*, returned when `output_hidden_states=True` is passed or when `config.output_hidden_states=True`): Tuple of `torch.FloatTensor` (one for the output of the embeddings, if the model has an embedding layer, + one for the output of each layer) of shape `(batch_size, sequence_length, hidden_size)`. Hidden-states of the model at the output of each layer plus the optional initial embedding outputs. attentions (`tuple(torch.FloatTensor)`, *optional*, returned when `output_attentions=True` is passed or when `config.output_attentions=True`): Tuple of `torch.FloatTensor` (one for each layer) of shape `(batch_size, num_heads, sequence_length, sequence_length)`. Attentions weights after the attention softmax, used to compute the weighted average in the self-attention heads. Nembedslast_hidden_state pooler_output hidden_states attentions)__name__ __module__ __qualname____firstlineno____doc__rgrr' FloatTensor__annotations__rhrirjrrk__static_attributes__rlr,r*reres}*+/FHU&& '.59x 1 12915M8E--.58