U 9%eK@sddlZddlZddlmZddlmZddlmZddl m Z e e Z dZ GdddZGdd d ZGd d d eZGd d d eZGdddeZGdddeZGdddeZGdddeZGdddeZGdddeZGdddeZGdddeZGdddeZdS) N)add_start_docstrings) get_loggerad Args: input_ids (`jnp.ndarray` of shape `(batch_size, sequence_length)`): Indices of input sequence tokens in the vocabulary. Indices can be obtained using [`PreTrainedTokenizer`]. See [`PreTrainedTokenizer.encode`] and [`PreTrainedTokenizer.__call__`] for details. [What are input IDs?](../glossary#input-ids) scores (`jnp.ndarray` of shape `(batch_size, config.vocab_size)`): Prediction scores of a language modeling head. These can be logits for each vocabulary when not using beam search or log softmax for each vocabulary token when using beam search kwargs (`Dict[str, Any]`, *optional*): Additional logits processor specific kwargs. Return: `jnp.ndarray` of shape `(batch_size, config.vocab_size)`: The processed prediction scores. c@s0eZdZdZeeejejejdddZdS)FlaxLogitsProcessorzSAbstract base class for all logit processors that can be applied during generation. input_idsscoresreturncCst|jddS)z"Flax method for processing logits.H is an abstract class. Only classes inheriting this class can be called.NNotImplementedError __class__selfrrrj/var/www/html/Darija-Ai-API/env/lib/python3.8/site-packages/transformers/generation/flax_logits_process.py__call__5s zFlaxLogitsProcessor.__call__N __name__ __module__ __qualname____doc__r!LOGITS_PROCESSOR_INPUTS_DOCSTRINGjnpndarrayrrrrrr2src@s0eZdZdZeeejejejdddZdS)FlaxLogitsWarperzjAbstract base class for all logit warpers that can be applied during generation with multinomial sampling.rcCst|jddS)zFlax method for warping logits.r Nr rrrrr@s zFlaxLogitsWarper.__call__Nrrrrrr=src@s2eZdZdZeeejejeejdddZ dS)FlaxLogitsProcessorLista. This class can be used to create a list of [`FlaxLogitsProcessor`] or [`FlaxLogitsWarper`] to subsequently process a `scores` input tensor. This class inherits from list and adds a specific *__call__* method to apply each [`FlaxLogitsProcessor`] or [`FlaxLogitsWarper`] to the inputs. rrcur_lenr c s|D]}t|jj}t|dkrztfddt|ddDshtdt|d|j d||||f}q||||}q|S)Nc3s|]}|kVqdSNr).0argkwargsrr Tsz3FlaxLogitsProcessorList.__call__..rz,Make sure that all the required parameters: z for z$ are passed to the logits processor.) inspect signaturer parameterslenalllistkeys ValueErrorr )rrrrr$ processorZ function_argsrr#rrOs &z FlaxLogitsProcessorList.__call__N) rrrrrrrrintrrrrrrHsrc@s8eZdZdZedddZejejeejdddZ dS) FlaxTemperatureLogitsWarperz [`FlaxLogitsWarper`] for temperature (exponential scaling output probability distribution). Args: temperature (`float`): The value used to module the logits distribution.  temperaturecCs*t|tr|dks td|||_dS)Nrz:`temperature` has to be a strictly positive float, but is ) isinstancefloatr-r2)rr2rrr__init__hsz$FlaxTemperatureLogitsWarper.__init__rcCs||j}|Sr r1rrrrrrrrns z$FlaxTemperatureLogitsWarper.__call__N) rrrrr4r5rrr/rrrrrr0_sr0c@sHeZdZdZed dfeeedddZejejeejddd Z d S) FlaxTopPLogitsWarperaH [`FlaxLogitsWarper`] that performs top-p, i.e. restricting to top tokens summing to prob_cut_off <= prob_cut_off. Args: top_p (`float`): If set to < 1, only the smallest set of most probable tokens with probabilities that add up to `top_p` or higher are kept for generation. filter_value (`float`, *optional*, defaults to `-float("Inf")`): All filtered values will be set to this float value. min_tokens_to_keep (`int`, *optional*, defaults to 1): Minimum number of tokens that cannot be filtered. Inf)top_p filter_valuemin_tokens_to_keepcCs^t|tr|dks|dkr(td|t|tr:|dkrHtd|||_||_||_dS)Nrg?z.`top_p` has to be a float > 0 and < 1, but is r9z:`min_tokens_to_keep` has to be a positive integer, but is )r3r4r-r/r:r;r<)rr:r;r<rrrr5szFlaxTopPLogitsWarper.__init__rc Cst||jd\}}t||j}tjj|ddj dd}||j k}t |d}||j dddf dO}|j ddd|jf d}t|||} tj|| d} | S)NZaxisr9rT)laxtop_kshaperZ full_liker;jaxnnZsoftmaxZcumsumr:Zrollatsetr<whereZ sort_key_val) rrrr topk_scores topk_indicesZ mask_scoresZcumulative_probsZ score_maskZtopk_next_scores next_scoresrrrrs  zFlaxTopPLogitsWarper.__call__N rrrrr4r/r5rrrrrrrr7ss  r7c@sHeZdZdZed dfeeedddZejejeejddd Z d S) FlaxTopKLogitsWarpera [`FlaxLogitsWarper`] that performs top-k, i.e. restricting to the k highest probability elements. Args: top_k (`int`): The number of highest probability vocabulary tokens to keep for top-k-filtering. filter_value (`float`, *optional*, defaults to `-float("Inf")`): All filtered values will be set to this float value. min_tokens_to_keep (`int`, *optional*, defaults to 1): Minimum number of tokens that cannot be filtered. r8r9)r@r;r<cCs6t|tr|dkr td|t|||_||_dS)Nrz6`top_k` has to be a strictly positive integer, but is )r3r/r-maxr@r;)rr@r;r<rrrr5s zFlaxTopKLogitsWarper.__init__rcCs|j\}}t|||j}t|j|jd}t||\}} tt||dddf||f } | } | | } |j |  | }| ||} | S)Nr=) rArfullr;minr@r?Z broadcast_toZarangeflattenrDrEZreshape)rrrr batch_size vocab_sizeZnext_scores_flatZtopkrGrHshiftZtopk_scores_flatZtopk_indices_flatrIrrrrs *  zFlaxTopKLogitsWarper.__call__NrJrrrrrKs rKc@s8eZdZdZedddZejejeejdddZdS) !FlaxForcedBOSTokenLogitsProcessorz [`FlaxLogitsProcessor`] that enforces the specified token as the first generated token. Args: bos_token_id (`int`): The id of the token to force as the first generated token.  bos_token_idcCs ||_dSr rT)rrUrrrr5sz*FlaxForcedBOSTokenLogitsProcessor.__init__rcCsNt|jtd }dt|d}t||jdd|jfd|}|SNinfr9r) rrMrAr4bool_rFrDrUrErrrr new_scores apply_penaltyrrrrs$z*FlaxForcedBOSTokenLogitsProcessor.__call__N rrrrr/r5rrrrrrrrSsrSc@s:eZdZdZeedddZejejeejdddZdS) !FlaxForcedEOSTokenLogitsProcessorae [`FlaxLogitsProcessor`] that enforces the specified token as the last generated token when `max_length` is reached. Args: max_length (`int`): The maximum length of the sequence to be generated. eos_token_id (`int`): The id of the token to force as the last generated token when `max_length` is reached.  max_length eos_token_idcCs||_||_dSr r^)rr_r`rrrr5sz*FlaxForcedEOSTokenLogitsProcessor.__init__rcCsTt|jtd }dt||jd}t||jdd|jf d|}|SrV) rrMrAr4rXr_rFrDr`rErYrrrrs$z*FlaxForcedEOSTokenLogitsProcessor.__call__Nr\rrrrr]s r]c@s:eZdZdZeedddZejejeejdddZdS) FlaxMinLengthLogitsProcessora3 [`FlaxLogitsProcessor`] enforcing a min-length by setting EOS probability to 0. Args: min_length (`int`): The minimum length below which the score of `eos_token_id` is set to `-float("Inf")`. eos_token_id (`int`): The id of the *end-of-sequence* token. ) min_lengthr`cCsPt|tr|dkr td|t|tr2|dkr@td|||_||_dS)Nrz2`min_length` has to be a positive integer, but is z4`eos_token_id` has to be a positive integer, but is )r3r/r-rbr`)rrbr`rrrr5s z%FlaxMinLengthLogitsProcessor.__init__rcCsFdt||jdd}t||jdd|jftd |}|S)Nr9rrW)rZcliprbrFrDr`rEr4rrrrr[rrrrs*z%FlaxMinLengthLogitsProcessor.__call__Nr\rrrrras  rac@s&eZdZdZddZedddZdS)(FlaxSuppressTokensAtBeginLogitsProcessora [`FlaxLogitsProcessor`] supressing a list of tokens as soon as the `generate` function starts generating using `begin_index` tokens. This should ensure that the tokens defined by `begin_suppress_tokens` are not sampled at the begining of the generation. Args: begin_suppress_tokens (`List[int]`): Tokens to not sample. begin_index (`int`): Index where the tokens are suppressed. cCst||_||_dSr )r+begin_suppress_tokens begin_index)rrerfrrrr5s z1FlaxSuppressTokensAtBeginLogitsProcessor.__init__)rcCsBdt||j}t||jdd|jftd |}|S)Nr9rW)rrXrfrFrDrerEr4rcrrrr s*z1FlaxSuppressTokensAtBeginLogitsProcessor.__call__N)rrrrr5r/rrrrrrds rdc@s8eZdZdZedddZejejeejdddZ dS) !FlaxSuppressTokensLogitsProcessorz [`FlaxLogitsProcessor`] suppressing a list of tokens at each decoding step. The processor will set their log probs to be `-inf` so they are not sampled. Args: suppress_tokens (`list`): Tokens to not sample. )suppress_tokenscCst||_dSr )r+rh)rrhrrrr52sz*FlaxSuppressTokensLogitsProcessor.__init__rcCs |jd|jftd }|S)N.rW)rDrhrEr4r6rrrr5sz*FlaxSuppressTokensLogitsProcessor.__call__N) rrrrr+r5rrr/rrrrrrg(s rgc@s2eZdZdZddZejejeejdddZdS)FlaxForceTokensLogitsProcessora [`FlaxLogitsProcessor`] that takes a list of pairs of integers which indicates a mapping from generation indices to token indices that will be forced before sampling. The processor will set their log probs to 0 and all other tokens to `-inf` so that they are sampled at their corresponding index. Args: force_token_map (`list`): Map giving token ids and indices where they will be forced to be sampled. cCsbt|}tjt|dtjdd}|D] \}}|dk r0|j||}q0t||_ dS)Nr9dtyper=) dictrZonesrLr,Zint32itemsrDrEforce_token_array)rZforce_token_maprnindextokenrrrr5Fs  z'FlaxForceTokensLogitsProcessor.__init__rcsBfddtjjdkfddfddS)NcsXjd}j|}tjjdtd }tj|dfjd}t||d|f}|S)NrrjrWr9) rArnrZ ones_likerkr4Zzerosr?Zdynamic_update_slice)Zgeneration_idxrP current_tokenrZZupdates)rrrr _force_tokenRs   z=FlaxForceTokensLogitsProcessor.__call__.._force_tokenrcsSr rrrrr^z9FlaxForceTokensLogitsProcessor.__call__..cs*tjdkfddfddS)NrcsSr rr)rrrrrrtcruzKFlaxForceTokensLogitsProcessor.__call__....csSr rrrsrrrteru)r?condrnrrrrrrrrrt`s   )r?rvrnrAr6rrwrrQs   z'FlaxForceTokensLogitsProcessor.__call__N) rrrrr5rrr/rrrrrri;s  ric@s eZdZdZddZddZdS)#FlaxWhisperTimeStampLogitsProcessora{ Whisper specific Processor. This processor can be used to force a list of tokens. The processor will set their log probs to `inf` so that they are sampled at their corresponding index. Args: generate_config (`GenerateConfig`): The generate config used to generate the output. The following parameters are required: eos_token_id (`int`, *optional*, defaults to 50257): The id of the *end-of-sequence* token. no_timestamps_token_id (`int`, *optional*, defaults to 50363): The id of the `"<|notimestamps|>"` token. max_initial_timestamp_index (`int`, *optional*, defaults to 1): Used to set the maximum value of the initial timestamp. This is used to prevent the model from predicting timestamps that are too far in the future. cCsl|j|_|j|_|jd|_|d|_|jr:|jd7_t|drN|j|_n|j|_|jdkrh|j|_dS)Nr9rmax_initial_timestamp_index)r`no_timestamps_token_idtimestamp_beginrfZis_multilingualhasattrryrQ)rZgenerate_configZ model_configZdecoder_input_lengthrrrr5|s     z,FlaxWhisperTimeStampLogitsProcessor.__init__c s|jddjftd }fdd}t|||}tjkdd}tj dk dod|d}j j }t||jdd|ddftd |}tj j |dd}fd d }t|||}|S) NrWc stjdkdd}t|djkdo2|d}tjdkdd}t|djkd|}t|t|dk|jjdtd |jdjtd |S)Nr9TFrrrW)rrFrfr{rDrEr4r`)Z input_ids_kscores_kZlast_was_timestampZpenultimate_was_timestamprrrr handle_pairss*zBFlaxWhisperTimeStampLogitsProcessor.__call__..handle_pairsTFr9r=r>csXtjj|jddd}t|dj}t||k|jdjt d |S)Nr=r>rW) rBrCZ logsumexpr{rrLrFrDrEr4)Z logprobs_kr}Ztimestamp_logprobZmax_text_token_logprob)rrrhandle_cumulative_probsszMFlaxWhisperTimeStampLogitsProcessor.__call__..handle_cumulative_probs) rDrzrEr4rBZvmaprrFrfryr{rCZ log_softmax) rrrrrZapply_max_initial_timestampZ last_allowedZlogprobsrrr~rrs&  $ z,FlaxWhisperTimeStampLogitsProcessor.__call__N)rrrrr5rrrrrrxksrx)r&rBZjax.laxr?Z jax.numpynumpyrutilsrZ utils.loggingrrloggerrrrr+rr0r7rKrSr]rardrgrirxrrrrs(      ,#0