fThSSKJrJrJrJr SSKrSSKJr SSKJ r SSK J r J r J r SSKJr SSKJr SSKJrJr SS KJr SS KJr SS KJrJrJrJrJr SS KJ r SS K!J"r"J#r# SSK$J%r% SSK&J'r'J(r(J)r) SSK*J+r+J,r,J-r- SSK.J/r/J0r0J1r1 SSK2J3r3J4r4 \)Rj"\65r7"SS\5r8"SS\Rr5r:"SS\Rr5r;"SS\+5r<"SS\,5r="SS\/5r>"S S!\5r?\'"S"S#\#55r@"S$S%\@5rA"S&S'\05rB"S(S)\35rC\'"S*S+9"S,S-\@\55rD/S.QrEg)/)CallableOptionalTupleUnionN)ACT2FN)Cache DynamicCacheEncoderDecoderCache)PretrainedConfig)GenerationMixin)_prepare_4d_attention_mask#_prepare_4d_attention_mask_for_sdpa)FlashAttentionKwargs)GradientCheckpointingLayer)BaseModelOutputBaseModelOutputWithPast)BaseModelOutputWithPastAndCrossAttentionsSeq2SeqLMOutputSeq2SeqModelOutput)rope_config_validation)ALL_ATTENTION_FUNCTIONSPreTrainedModel)Unpack)auto_docstringcan_return_tuplelogging) GlmAttentionGlmRotaryEmbeddingapply_rotary_pos_emb)LlamaDecoderLayer LlamaModeleager_attention_forward) WhisperModelshift_tokens_rightcv^\rSrSrSrSrS/rSSSS.rS U4S jjrS r U=r $) MoonshineConfig.a! This is the configuration class to store the configuration of a [`MoonshineModel`]. It is used to instantiate a Moonshine model according to the specified arguments, defining the model architecture. Instantiating a configuration with the defaults will yield a similar configuration to that of the Moonshine [UsefulSensors/moonshine-tiny](https://huggingface.co/UsefulSensors/moonshine-tiny). Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the documentation from [`PretrainedConfig`] for more information. Args: vocab_size (`int`, *optional*, defaults to 32768): Vocabulary size of the Moonshine model. Defines the number of different tokens that can be represented by the `inputs_ids` passed when calling [`MoonshineModel`]. hidden_size (`int`, *optional*, defaults to 288): Dimension of the hidden representations. intermediate_size (`int`, *optional*, defaults to 1152): Dimension of the MLP representations. encoder_num_hidden_layers (`int`, *optional*, defaults to 6): Number of hidden layers in the Transformer encoder. decoder_num_hidden_layers (`int`, *optional*, defaults to 6): Number of hidden layers in the Transformer decoder. encoder_num_attention_heads (`int`, *optional*, defaults to 8): Number of attention heads for each attention layer in the Transformer encoder. decoder_num_attention_heads (`int`, *optional*, defaults to 8): Number of attention heads for each attention layer in the Transformer decoder. encoder_num_key_value_heads (`int`, *optional*): This is the number of key_value heads that should be used to implement Grouped Query Attention. If `encoder_num_key_value_heads=encoder_num_attention_heads`, the model will use Multi Head Attention (MHA), if `encoder_num_key_value_heads=1` the model will use Multi Query Attention (MQA) otherwise GQA is used. When converting a multi-head checkpoint to a GQA checkpoint, each group key and value head should be constructed by meanpooling all the original heads within that group. For more details checkout [this paper](https://arxiv.org/pdf/2305.13245.pdf). If it is not specified, will default to `num_attention_heads`. decoder_num_key_value_heads (`int`, *optional*): This is the number of key_value heads that should be used to implement Grouped Query Attention. If `decoder_num_key_value_heads=decoder_num_attention_heads`, the model will use Multi Head Attention (MHA), if `decoder_num_key_value_heads=1` the model will use Multi Query Attention (MQA) otherwise GQA is used. When converting a multi-head checkpoint to a GQA checkpoint, each group key and value head should be constructed by meanpooling all the original heads within that group. For more details checkout [this paper](https://arxiv.org/pdf/2305.13245.pdf). If it is not specified, will default to `decoder_num_attention_heads`. pad_head_dim_to_multiple_of (`int`, *optional*): Pad head dimension in encoder and decoder to the next multiple of this value. Necessary for using certain optimized attention implementations. encoder_hidden_act (`str` or `function`, *optional*, defaults to `"gelu"`): The non-linear activation function (function or string) in the encoder. decoder_hidden_act (`str` or `function`, *optional*, defaults to `"silu"`): The non-linear activation function (function or string) in the decoder. max_position_embeddings (`int`, *optional*, defaults to 512): The maximum sequence length that this model might ever be used with. initializer_range (`float`, *optional*, defaults to 0.02): The standard deviation of the truncated_normal_initializer for initializing all weight matrices. decoder_start_token_id (`int`, *optional*, defaults to 1): Corresponds to the "<|startoftranscript|>" token, which is automatically used when no `decoder_input_ids` are provided to the `generate` function. It is used to guide the model`s generation process depending on the task. use_cache (`bool`, *optional*, defaults to `True`): Whether or not the model should return the last key/values attentions (not used by all models). rope_theta (`float`, *optional*, defaults to 10000.0): The base period of the RoPE embeddings. rope_scaling (`Dict`, *optional*): Dictionary containing the scaling configuration for the RoPE embeddings. NOTE: if you apply new rope type and you expect the model to work on longer `max_position_embeddings`, we recommend you to update this value accordingly. Expected contents: `rope_type` (`str`): The sub-variant of RoPE to use. Can be one of ['default', 'linear', 'dynamic', 'yarn', 'longrope', 'llama3'], with 'default' being the original RoPE implementation. `factor` (`float`, *optional*): Used with all rope types except 'default'. The scaling factor to apply to the RoPE embeddings. In most scaling types, a `factor` of x will enable the model to handle sequences of length x * original maximum pre-trained length. `original_max_position_embeddings` (`int`, *optional*): Used with 'dynamic', 'longrope' and 'llama3'. The original max position embeddings used during pretraining. `attention_factor` (`float`, *optional*): Used with 'yarn' and 'longrope'. The scaling factor to be applied on the attention computation. If unspecified, it defaults to value recommended by the implementation, using the `factor` field to infer the suggested value. `beta_fast` (`float`, *optional*): Only used with 'yarn'. Parameter to set the boundary for extrapolation (only) in the linear ramp function. If unspecified, it defaults to 32. `beta_slow` (`float`, *optional*): Only used with 'yarn'. Parameter to set the boundary for interpolation (only) in the linear ramp function. If unspecified, it defaults to 1. `short_factor` (`List[float]`, *optional*): Only used with 'longrope'. The scaling factor to be applied to short contexts (< `original_max_position_embeddings`). Must be a list of numbers with the same length as the hidden size divided by the number of attention heads divided by 2 `long_factor` (`List[float]`, *optional*): Only used with 'longrope'. The scaling factor to be applied to long contexts (< `original_max_position_embeddings`). Must be a list of numbers with the same length as the hidden size divided by the number of attention heads divided by 2 `low_freq_factor` (`float`, *optional*): Only used with 'llama3'. Scaling factor applied to low frequency components of the RoPE `high_freq_factor` (`float`, *optional*): Only used with 'llama3'. Scaling factor applied to high frequency components of the RoPE partial_rotary_factor (`float`, *optional*, defaults to 0.9): Percentage of the query and keys which will have rotary embedding. is_encoder_decoder (`bool`, *optional*, defaults to `True`): Whether the model is used as an encoder/decoder or not. attention_bias (`bool`, *optional*, defaults to `False`): Whether to use a bias in the query, key, value and output projection layers during self-attention. attention_dropout (`float`, *optional*, defaults to 0.0): The dropout ratio for the attention probabilities. bos_token_id (`int`, *optional*, defaults to 1): Denotes beginning of sequences token id. eos_token_id (`int`, *optional*, defaults to 2): Denotes end of sequences token id. Example: ```python >>> from transformers import MoonshineModel, MoonshineConfig >>> # Initializing a Moonshine style configuration >>> configuration = MoonshineConfig().from_pretrained("UsefulSensors/moonshine-tiny") >>> # Initializing a model from the configuration >>> model = MoonshineModel(configuration) >>> # Accessing the model configuration >>> configuration = model.config ``` moonshinepast_key_valuesencoder_num_key_value_headsencoder_num_attention_headsencoder_num_hidden_layers)num_key_value_headsnum_attention_headsnum_hidden_layersc p>XlX lX0lX@lXPlX`lXplUcUnXlU cUn XlXl Xl Xl Xl Xl XlUUlUUlUUlUUlUUlUUlUUl[-U5 [.TU]`"SUUUUS.UD6 g)N) bos_token_id eos_token_idis_encoder_decoderdecoder_start_token_id) vocab_size hidden_sizeintermediate_sizer.decoder_num_hidden_layersr-decoder_num_attention_headsr,decoder_num_key_value_headspad_head_dim_to_multiple_ofencoder_hidden_actdecoder_hidden_actmax_position_embeddingsinitializer_ranger6 use_cache rope_theta rope_scalingpartial_rotary_factorr5attention_biasattention_dropoutrsuper__init__)selfr8r9r:r.r;r-r<r,r=r>r?r@rArBr6rCrDrErFr5rGrHr3r4kwargs __class__s g/var/www/auris/envauris/lib/python3.13/site-packages/transformers/models/moonshine/modular_moonshine.pyrJMoonshineConfig.__init__s8%&!2)B&)B&+F(+F( & .*E '+F( & .*E '+F(+F("4"4'>$!2&<#"$(%:""4,!2 t$  %%1#9    )rGrHr@r<r;r=r6r?r-r.r,r9rBr:r5rAr>rFrErDrCr8)ii irQrRNNNgelusiluig{Gz?Tg@Ng?TFrUr) __name__ __module__ __qualname____firstlineno____doc__ model_typekeys_to_ignore_at_inference attribute_maprJ__static_attributes__ __classcell__rMs@rNr(r(.s{ zJ#4"5<<8M"#"#$%$%$($($(!! # !3D D rPr(cb^\rSrSrU4SjrS\R S\R 4SjrSrU=r $)MoonshineEncoderMLPc >[TU]5 Xl[UUl[ R "URUR5Ul [ R "URUR5Ul gN rIrJconfigr activation_fnnnLinearr9r:fc1fc2rKrh hidden_actrMs rNrJMoonshineEncoderMLP.__init__s\  #J/99V//1I1IJ99V55v7I7IJrP hidden_statesreturnclURU5nURU5nURU5nU$rf)rlrirm)rKrqs rNforwardMoonshineEncoderMLP.forwards4/ **=9 / rPrirhrlrm rWrXrYrZrJtorchTensorrtr_r`ras@rNrcrcs)KU\\ellrPrccb^\rSrSrU4SjrS\R S\R 4SjrSrU=r $)MoonshineDecoderMLPi c>[TU]5 Xl[UUl[ R "URURS-5Ul [ R "URUR5Ul g)Nrrgrns rNrJMoonshineDecoderMLP.__init__ sa  #J/99V//1I1IA1MN99V55v7I7IJrPrqrrcURU5nURSSS9upURU5U-nURU5nU$)Nr)dim)rlchunkrirm)rKrqgates rNrtMoonshineDecoderMLP.forwardsQ/ +11!1< **40=@ / rPrvrwras@rNr{r{ s)KU\\ellrPr{cx^\rSrSrS\S\S\S\S\4 U4SjjrSS\RS \ \ \R\R4S \ \RS \ \ S \ \RS \ \RS\\S\ \R\ \R\ \ \R44SjjrSrU=r$)MoonshineAttentionirh layer_idx is_causalr0r/cf>URXES.5 [TU] X5 X0l[ USUR UR -5UlURRbAURRnX`RU-S- U--nXpR- Ul gSUl g)N)r0r/head_dimrUr) updaterIrJrgetattrr9r0rrhr>head_dim_padding) rKrhrrr0r/target_multipletarget_head_dimrMs rNrJMoonshineAttention.__init__s  .Ano +" F4F4F&JdJd4de  ;; 2 2 >"kkEEO---/2QTU2UZi1ijO$3mm$CD !$%D !rPrqposition_embeddingsattention_maskpast_key_valuecache_positionkey_value_statesrLrrc URSSupURU5RXURRUR 5R SS5n USLn Ub^URRUR5n U (a&SURUR'URnO URnUbUOUn U (aAU(a:W (a3URURnURURnOURU 5RUSURRUR 5R SS5nURU 5RUSURRUR 5R SS5nU (a$Ub!UR!XURSU05upU (d<Uunn[#XUU5upUb%UUUS.nUR!XURU5up[$nURR&S:wadURR&S:Xa-URS S 5(a[(R+S 5 O[,URR&nUR.(a UcU S:aSOS nUR0S :a[2R4R6R9U S UR045n [2R4R6R9US UR045n[2R4R6R9US UR045nU"UU UUU4UR:(dS O UR<UR>US.UD6unnUR0S :aUSSUR0*24nURAXS5RC5nUREU5nUU4$)NrrUrTr)sincosreagersdpaoutput_attentionsFz`torch.nn.functional.scaled_dot_product_attention` does not support `output_attentions=True`. Falling back to eager attention. This warning can be removed using the argument `attn_implementation="eager"` when loading the model.rrV)dropoutscalingr.)#shapeq_projviewrhr/r transpose is_updatedgetrcross_attention_cacheself_attention_cache key_cache value_cachek_projv_projrr!r$_attn_implementationlogger warning_oncerrrrxrj functionalpadtrainingrHrreshape contiguouso_proj)rKrqrrrrrrLbszq_len query_statesis_cross_attentionrcurrent_states key_states value_statesrr cache_kwargsattention_interfacer attn_output attn_weightss rNrtMoonshineAttention.forward0s#(("-  KK & + +C 8W8WY]YfYf g q qrsuv w .T9  %'2266t~~FJ!<@))$..9!/!E!E!/!D!D.>-I)} .Z'11$..AJ)55dnnEL N+c2t{{>> N1a   N+c2t{{>> N1a  "n&@+9+@+@dnn?OQ_>`,( "*HC';LVY[^'_ $L)'*3.Y +9+@+@dnnl,( )@ ;; + +w 6{{//69fjjI\^c>d>d##L '>dkk>^>^&_# NN~/E%RS)DY^  1 $ 88..22Q5RR S[)[)rPrc\rSrSrSrg)MoonshineRotaryEmbeddingir7N)rWrXrYrZr_r7rPrNrrsrPrc4^\rSrSrS\S\4U4SjjrSrU=r$)MoonshineEncoderLayerirhrc4>[TU]X5 [UUSURURS9Ul[ XR5Ul[R"URSS9Ul [R"URSS9Ul g)NFrhrrr0r/bias)rIrJrr-r, self_attnrcr?mlprj LayerNormr9input_layernormpost_attention_layernormrKrhrrMs rNrJMoonshineEncoderLayer.__init__s} ++ & B B & B B  'v/H/HI!||F,>,>UK(* V5G5Ge(T%rP)rrrr) rWrXrYrZr(rrJr_r`ras@rNrrs U U3 U UrPrc ^\rSrSrSS\S\\4U4SjjjrSS\RS\\RS\\RS\\RS \\RS \\RS \\ S \\ S \\ S\\RS\\ \R\R4S\\ \R\R4S\ \R\\ \R\R444SjjrSrU=r$)MoonshineDecoderLayerirhrc>[TU]5 URUl[UUSURUR S9Ul[UUSURUR S9Ul[XR5Ul [R"URSS9Ul [R"URSS9Ul[R"URSS9Ulg)NTrFr)rIrJr9rr<r=r encoder_attnr{r@rrjrrrfinal_layernormrs rNrJMoonshineDecoderLayer.__init__s !--+ & B B & B B  / & B B & B B  'v/H/HI!||F,>,>UK(* V5G5Ge(T%!||F,>,>UKrPrqrencoder_hidden_statesencoder_attention_mask position_idsencoder_position_idsrrrCrrencoder_position_embeddingsrrc FUnURU5nUR"SUUUUUU U U S.U D6upX-nSnUb.UnURU5nURUUUUUU S9unnX-nUnUR U5nUR U5nX-nU4nU(aUUU4- nU$)N)rqrrrrrCrr)rqrrrrrCr7)rrrrrr)rKrqrrrrrrrrCrrrrLresidualself_attn_weightscross_attn_weightsoutputss rNrtMoonshineDecoderLayer.forwards !,,]; ,0>> , ')%)/) 3 ,  , ( !0 " ,$H 99-HM040A0A+!65-"3# 1B1 -M-%4M!,,]; /  0  "  )+=> >GrP)rrr9rrrrrf) NNNNNNFFNNN)rWrXrYrZr(rrrJrxryrr rr FloatTensorrtr_r`ras@rNrrsuLL8C=LL6268<9=37;?*.,1$)59KOSW<||<!.< ( 5 < !) 6 < u//0 <'u'7'78<!<$D><D><!!1!12<&eELL%,,,F&GH<&.eELL%,,4N.O%P< u  (51B1BEDUDU1U+V"WW X<<rPrcd\rSrSr\rSrSrSrSS/r Sr Sr Sr Sr SrS\R 4S jrS rg ) MoonshinePreTrainedModelimodel input_valuesTrrcPURRn[U[R[R 45(aWUR RRSUS9 URb%URRR5 gg[U[R[R45(aXUR RRS5 URb%URRR5 gg[U[R5(adUR RRSUS9 URb2UR RURR5 ggg)NrV)meanstdg?)rhrB isinstancerjrkConv1dweightdatanormal_rzero_ GroupNormrfill_ Embedding padding_idx)rKmodulers rN _init_weights&MoonshinePreTrainedModel._init_weightss)kk++ fryy"))4 5 5 MM   & &CS & 9{{&   &&(' r|| < = = MM   $ $S ){{&   &&('  - - MM   & &CS & 9!!- ""6#5#56<<>..rP input_lengthsc~[US- S- S-5n[US- S- S-5n[US- S- S-5nU$)z8 Computes the output length of the convolutional layers @rUrr)r)rKroutput_conv1_lengthoutput_conv2_lengthoutput_conv3_lengths rN _get_feat_extract_output_lengths9MoonshinePreTrainedModel._get_feat_extract_output_lengthssZ"=3#6""N>N#rPrc^\rSrSrSrSrS\4U4SjjrS\R4Sjr S\R4S jr \ SS\ \RS \ \R S \ \S \ \S \\S\4 Sjj5rSrU=r$)MoonshineEncoderi!z Transformer encoder consisting of *config.num_hidden_layers* layers. Each layer is a [`MoonshineEncoderLayer`] Args: config: MoonshineConfig rrhc L>[TU]U5 XlURn[R "SUSSSS9Ul[R "USU-SSS 9Ul[R "SU-USSS 9Ul[R"SUS S 9Ul [US 9Ul [R"[UR5Vs/sHn[!X5PM sn5Ul[R$"USS 9UlSUlUR+5 gs snf)NrUrrF) kernel_sizestriderrrr)r r gh㈵>) num_groups num_channelseps)rhr)rIrJrhr9rjrconv1conv2conv3r groupnormr rotary_emb ModuleListranger.rlayersr layer_normgradient_checkpointing post_init)rKrh embed_dimidxrMs rNrJMoonshineEncoder.__init__+s   && YYq)ReT YYy!i-QqQ YYq9}iQqQ PTU2&Amm;@AaAa;b c;bC "6 /;b c  ,,yu=&+#  dsD!rrcUR$rfrrKs rNget_input_embeddings%MoonshineEncoder.get_input_embeddings?s zzrPvaluecXlgrfr )rKr$s rNset_input_embeddings%MoonshineEncoder.set_input_embeddingsBs rPrroutput_hidden_statesflash_attn_kwargsc UbUOURRnUbUOURRnUc [S5eUR S5n[ R RURU55nURU5n[ R RURU55n[ R RURU55nURSSS5nUbURURS5nSnUSSSU24SSU24nURR S :XaUS :HR#5(aUOSnOLURR S :XaU(d[%X&R&5nO[)X&R&5n[*R,"SURSUR.S 9R S5n UR1Xi5n U(aS OSn U(aS OSn UR2H3n U(aX4- n U "U4UU UU S.UD6nUSnU(dM+XS4- n M5 UR5U5nU(aX4- n [7UU U S9$)a\ Args: input_values (`torch.FloatTensor` of shape `(batch_size, audio_length)`): Float values of the raw speech waveform. Raw speech waveform can be obtained by loading a `.flac` or `.wav` audio file into an array of type `List[float]` or a `numpy.ndarray`, *e.g.* via the soundfile library (`pip install soundfile`). To prepare the array into `input_values`, the [`AutoFeatureExtractor`] should be used for padding and conversion into a tensor of type `torch.FloatTensor`. attention_mask (`torch.Tensor` of shape `(batch_size, sequence_length)`, *optional*): Mask to avoid performing attention on padding indices in `input_values`. Mask values selected in `[0, 1]`: - 1 for tokens that are **not masked**, - 0 for tokens that are **masked**. [What are attention masks?](../glossary#attention-mask) output_attentions (`bool`, *optional*): Whether or not to return the attentions tensors of all attention layers. See `attentions` under returned tensors for more detail. output_hidden_states (`bool`, *optional*): Whether or not to return the hidden states of all layers. See `hidden_states` under returned tensors for more detail. return_dict (`bool`, *optional*): Whether or not to return a [`~utils.ModelOutput`] instead of a plain tuple. NzYou must specify input_values.rUrrr.flash_attention_2rVrdevicer7)rrrrlast_hidden_staterq attentions)rhrr( ValueError unsqueezerjrtanhrrrSrrpermuterrranyrdtyperrxaranger.rrrr)rKrrrr(r)rqmask_lendownsample_striderrall_hidden_statesall_self_attns encoder_layer layer_outputss rNrtMoonshineEncoder.forwardEs>2C1N-TXT_T_TqTq$8$D $++JjJj   => >$--a0  **4::l+CD }5  **4::m+DE  **4::m+DE %--aA6   %<<^=Q=QRT=UVH * +C1D3D1D,DEc9H9nUN{{//3FF4Bc4I3N3N3P3PVZ11V;DU!D^UhUh!i" c 'tn c $$89 c  !c c rPr cj^\rSrSrSrS\4U4SjjrSS\\RS\\RS\\RS\\ S\\RS \\ S \\ S \\ S \\RS \\RS\\RS\\S\\\44SjjrSrU=r$)MoonshineDecoderi input_idsrhc >[TU]U5 [R"URSS9Ul[R "[UR5Vs/sHn[X5PM sn5Ul gs snfNFr) rIrJrjrr9normrrr;rr)rKrhrrMs rNrJMoonshineDecoder.__init__s`  LL!3!3%@ mm;@AaAa;b c;bC "6 /;b c  csA>rrr+ inputs_embedsrCrr(rrrr)rrc  UbUOURRnUbUOURRnUbUOURRnUSLUSL- (a [ S5eUR (a/UR (aU(a[RS5 SnUcURU5nU(a"Uc[5n [5n[X5nU cDUbUR5OSn[R"XURS-UR S9n UcU R#S5nUR%X%XU5nUnUR'UU5nU(aSOSnU(aSOSnU(aU bSOSnU bU RS nS nU S SSU24S SU24n URR(S :XaU S :HR+5(aU OSn OjURR(S:Xa,U(d%[-U UR.URS 5n O$[1U UR.URS 5n UR2HHnU(aUU4- nU"U4UU U UUUUU US. U D6nUSnU(dM1UUS4- nU cM?UUS4- nMJ UR5U5nU(aUU4- n[7UU(aUOSUUUS9$)a\ encoder_hidden_states (`torch.FloatTensor` of shape `(batch_size, encoder_sequence_length, hidden_size)`, *optional*): Sequence of hidden-states at the output of the last layer of the encoder. Used in the cross-attention of the decoder. encoder_attention_mask (`torch.Tensor` of shape `(batch_size, sequence_length)`, *optional*): Mask to avoid performing attention on padding indices in `encoder_hidden_states`. Mask values selected in `[0, 1]`: - 1 for tokens that are **not masked**, - 0 for tokens that are **masked**. [What are attention masks?](../glossary#attention-mask) Nz:You must specify exactly one of input_ids or inputs_embedszX`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`.FrrUr-r7r+.r,rVr) rrrrrrrCrrr)r0r+rqr1cross_attentions)rhrr(rCr2rrrr embed_tokensr r get_seq_lengthrxr8rr.r3_update_causal_maskrrr6rr7rrrFr)rKrCrrr+rHrCrr(rrrr)rrpast_seen_tokens causal_maskrqrr;r<all_cross_attentionsr9r: decoder_layerr>s rNrtMoonshineDecoder.forwards;22C1N-TXT_T_TqTq$8$D $++JjJj "+!6IDKK $0N !12F^O  !CRC^==?de "\\ ]5H5H5K"KTaThThN  )33A6L.. >L] & #oom\J#7BD0d&7A $D>A 'tnA !!1!12A  ((9(9:A !) 6A $$89A  u-- .A A rPrBc\rSrSr\\SS\\RS\\RS\\RS\\RS\\ \ \RS\\ \ \ \R4S \\ \RS \\ \RS \\ S \\ S \\ S\\RS\4Sjj55rSrg)MoonshineModeli:Nrrdecoder_input_idsdecoder_attention_maskencoder_outputsr+decoder_inputs_embedsdecoder_position_idsrCrr(rrrc lU bU OURRn U bU OURRn U bU OURRn UcUR UUU U S9nOK[ U[ 5(d6[ US[U5S:aUSOS[U5S:aUSOSS9nURUUUURUUUU U U U S9 n [U RU RU RU RU RURURURS9$) a input_values (`torch.FloatTensor` of shape `(batch_size, audio_length)`): Float values of the raw speech waveform. Raw speech waveform can be obtained by loading a `.flac` or `.wav` audio file into an array of type `List[float]` or a `numpy.ndarray`, *e.g.* via the soundfile library (`pip install soundfile`). To prepare the array into `input_values`, the [`AutoFeatureExtractor`] should be used for padding and conversion into a tensor of type `torch.FloatTensor`. decoder_input_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`): Indices of input sequence tokens in the vocabulary. Padding will be ignored by default should you provide it. Indices can be obtained using [`AutoTokenizer`]. See [`PreTrainedTokenizer.encode`] and [`PreTrainedTokenizer.__call__`] for details. [What are input IDs?](../glossary#input-ids) decoder_attention_mask (`torch.Tensor` of shape `(batch_size, sequence_length)`, *optional*): Mask to avoid performing attention on padding token indices. Mask values selected in `[0, 1]`: - 1 for tokens that are **not masked**, - 0 for tokens that are **masked**. [What are attention masks?](../glossary#attention-mask) Indices can be obtained using [`AutoTokenizer`]. See [`PreTrainedTokenizer.encode`] and [`PreTrainedTokenizer.__call__`] for details. If `past_key_values` is used, optionally only the last `decoder_input_ids` have to be input (see `past_key_values`). If you want to change padding behavior, you should read [`modeling_opt._prepare_decoder_attention_mask`] and modify to your needs. See diagram 1 in [the paper](https://arxiv.org/abs/1910.13461) for more information on the default strategy. - 1 indicates the head is **not masked**, - 0 indicates the head is **masked**. decoder_inputs_embeds (`torch.FloatTensor` of shape `(batch_size, sequence_length, hidden_size)`, *optional*): Optionally, instead of passing `decoder_input_ids` you can choose to directly pass an embedded representation. This is useful if you want more control over how to convert `decoder_input_ids` indices into associated vectors than the model's internal embedding lookup matrix. decoder_position_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*): Indices of positions of each input sequence tokens in the position embeddings. Selected in the range `[0, config.n_positions - 1]`. [What are position IDs?](../glossary#position-ids) Example: ```python >>> import torch >>> from transformers import AutoFeatureExtractor, MoonshineModel >>> from datasets import load_dataset >>> model = MoonshineModel.from_pretrained("UsefulSensors/moonshine-tiny") >>> feature_extractor = AutoFeatureExtractor.from_pretrained("UsefulSensors/moonshine-tiny") >>> ds = load_dataset("hf-internal-testing/librispeech_asr_dummy", "clean", split="validation") >>> inputs = feature_extractor(ds[0]["audio"]["array"], return_tensors="pt") >>> input_values = inputs.input_values >>> decoder_input_ids = torch.tensor([[1, 1]]) * model.config.decoder_start_token_id >>> last_hidden_state = model(input_values, decoder_input_ids=decoder_input_ids).last_hidden_state >>> list(last_hidden_state.shape) [1, 2, 288] ``` N)rrr(rrUrr/) rCrrrr+rHrrCrr(r)r0r+decoder_hidden_statesdecoder_attentionsrKencoder_last_hidden_staterencoder_attentions)rhrr(rCencoderrrlendecoderr0rr+rqr1rK)rKrrrVrWrXr+rYrZrCrr(rdecoder_outputss rNrtMoonshineModel.forward;s^`2C1N-TXT_T_TqTq$8$D $++JjJj "+!6IDKK?{ D>{ $D>{ 'tn{ !!1!12{  { { rPrUzj The Moonshine Model with a language modeling head. Can be used for automatic speech recognition. ) custom_introc"^\rSrSrS/rS\4U4SjjrSrSrSr Sr S \ R4S jr \\SS \\R$S \\R&S \\R&S\\R&S\\\\R$S\\\\\R$4S\\\R$S\\\R&S\\S\\S\\S\\R&S\\R&S \4Sjj55rSrU=r$)!MoonshineForConditionalGenerationizproj_out.weightrhc>[TU]U5 [U5Ul[R "UR URSS9UlUR5 grE) rIrJrUrrjrkr9r8proj_outr)rKrhrMs rNrJ*MoonshineForConditionalGeneration.__init__sH  #F+  &"4"4f6G6GeT  rPc6URR5$rf)r get_encoderr!s rNrl-MoonshineForConditionalGeneration.get_encoderzz%%''rPc6URR5$rf)r get_decoderr!s rNrp-MoonshineForConditionalGeneration.get_decoderrnrPcUR$rfrir!s rNget_output_embeddings7MoonshineForConditionalGeneration.get_output_embeddingss }}rPcXlgrfrs)rKnew_embeddingss rNset_output_embeddings7MoonshineForConditionalGeneration.set_output_embeddingss& rPrrc6URR5$rf)rr"r!s rNr"6MoonshineForConditionalGeneration.get_input_embeddingsszz..00rPrrrVrWrXr+rYrZrCrr(rlabelscU b:Uc7Uc4[XRRURR5nUR UUUUUUUUU U U U S9 nUR UR 5nSnU b$URXURRS9n[UUURURURURURURUR S9 $)aw input_values (`torch.FloatTensor` of shape `(batch_size, audio_length)`): Float values of the raw speech waveform. Raw speech waveform can be obtained by loading a `.flac` or `.wav` audio file into an array of type `List[float]` or a `numpy.ndarray`, *e.g.* via the soundfile library (`pip install soundfile`). To prepare the array into `input_values`, the [`AutoFeatureExtractor`] should be used for padding and conversion into a tensor of type `torch.FloatTensor`. decoder_input_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`): Indices of input sequence tokens in the vocabulary. Padding will be ignored by default should you provide it. Indices can be obtained using [`AutoTokenizer`]. See [`PreTrainedTokenizer.encode`] and [`PreTrainedTokenizer.__call__`] for details. [What are input IDs?](../glossary#input-ids) decoder_attention_mask (`torch.Tensor` of shape `(batch_size, sequence_length)`, *optional*): Mask to avoid performing attention on padding token indices. Mask values selected in `[0, 1]`: - 1 for tokens that are **not masked**, - 0 for tokens that are **masked**. [What are attention masks?](../glossary#attention-mask) Indices can be obtained using [`AutoTokenizer`]. See [`PreTrainedTokenizer.encode`] and [`PreTrainedTokenizer.__call__`] for details. If `past_key_values` is used, optionally only the last `decoder_input_ids` have to be input (see `past_key_values`). If you want to change padding behavior, you should read [`modeling_opt._prepare_decoder_attention_mask`] and modify to your needs. See diagram 1 in [the paper](https://arxiv.org/abs/1910.13461) for more information on the default strategy. - 1 indicates the head is **not masked**, - 0 indicates the head is **masked**. decoder_inputs_embeds (`torch.FloatTensor` of shape `(batch_size, sequence_length, hidden_size)`, *optional*): Optionally, instead of passing `decoder_input_ids` you can choose to directly pass an embedded representation. This is useful if you want more control over how to convert `decoder_input_ids` indices into associated vectors than the model's internal embedding lookup matrix. decoder_position_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*): Indices of positions of each input sequence tokens in the position embeddings. Selected in the range `[0, config.n_positions - 1]`. [What are position IDs?](../glossary#position-ids) labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*): Labels for computing the language modeling loss. Indices should either be in `[0, ..., config.vocab_size]` or -100 (see `input_ids` docstring). Tokens with indices set to `-100` are ignored (masked), the loss is only computed for the tokens with labels in `[0, ..., config.vocab_size]`. Example: ```python >>> import torch >>> from transformers import AutoProcessor, MoonshineForConditionalGeneration >>> from datasets import load_dataset >>> processor = AutoProcessor.from_pretrained("UsefulSensors/moonshine-tiny") >>> model = MoonshineForConditionalGeneration.from_pretrained("UsefulSensors/moonshine-tiny") >>> ds = load_dataset("hf-internal-testing/librispeech_asr_dummy", "clean", split="validation") >>> inputs = processor(ds[0]["audio"]["array"], return_tensors="pt") >>> input_values = inputs.input_values >>> generated_ids = model.generate(input_values, max_new_tokens=100) >>> transcription = processor.batch_decode(generated_ids, skip_special_tokens=True)[0] >>> transcription 'Mr. Quilter is the apostle of the middle classes, and we are glad to welcome his gospel.' ```N) rrVrXrWr+rYrZrCrr(r)logitsr|r8) lossr~r+r\r]rKr^rr_)r&rh pad_token_idr6rrir0 loss_functionr8rr+r\r]rKr^rr_)rKrrrVrWrXr+rYrZrCrr(rr|rr~rs rNrt)MoonshineForConditionalGeneration.forwardsr   (-B-J$6KK44dkk6X6X%!'+jj )/+#9+"7!5/!5)'1 ' w889  %%Vt{{OeOe%fD#33")"?"?&99$55&-&G&G")"?"?&99  rP)rri) NNNNNNNNNNNNN)rWrXrYrZ_tied_weights_keysr(rJrlrprtrxrjr@r"rrrrxrrrrr rrrtr_r`ras@rNrgrgs ,,(('1bii159598<=AEIZ^DHBF$(,0/359-1{ u001{ !!1!12{ $E$4$45 { !))9)9 : { "%e.?.?(@"AB { "%(;U5CTCT=U(U"VW{  (e.?.?(@A{ 'uU-=-='>?{ D>{ $D>{ 'tn{ !!1!12{ ))*{  { { rPrg)r(rUrrg)Ftypingrrrrrxtorch.nnrj activationsr cache_utilsr r r configuration_utilsr generationr modeling_attn_mask_utilsrrmodeling_flash_attention_utilsrmodeling_layersrmodeling_outputsrrrrrmodeling_rope_utilsrmodeling_utilsrrprocessing_utilsrutilsrrrglm.modeling_glmrr r!llama.modeling_llamar"r#r$whisper.modeling_whisperr%r& get_loggerrWrr(r@rcr{rrrrrr rBrUrg__all__r7rPrNrsO43 !CC3)gB9:F&>>UUYYG   H %J &J Z "))  "))  q)q)h 1 U-U"U6Up"#"#"#JH /H VK zK \~ \~ B W (@/W  W t rP