tjdZddlmZddlZddlZddlZddlmZddlm Z m Z m Z ddl m Z ddlZddlZddlmZmZmZmZmZmZmZ ddlmZn #e$rdZYnwxYw dd lmZmZn#ee e!f$rdZdZYnwxYwej"e#Z$d d d d d Z%d Z&d Z'ddZ-d?d!Z.d@d$Z/dAd'Z0dBd*Z1Gd+d,Z2 dCdDd1Z3 dCdEd4Z4dFd6Z5dGd;Z6dS)HzAModality detection, input parsing, and message format conversion.) annotationsN) defaultdict)AnyLiteralNoReturn)urlparse)MULTIMODAL_DICT_KEYS AudioInput MessageFormatModality PairInput SingleInput VideoInput)Image) AudioDecoder VideoDecoderflat)apertus deepseek_v3gpt_ossseed_oss)querydocument))alphabravo)charlier)rfoxtrottextstrreturnboolc4|dod|vS)zQCheck if a string looks like a valid URL (starts with http(s) and has no spaces).)zhttp://zhttps:// ) startswithrs a/var/www/html/CA-Chatbot/venv/lib/python3.11/site-packages/sentence_transformers/base/modality.py_looks_like_urlr'7s ??2 3 3 G4G extensionstuple[str, ...]cHt|rN t|j}n#t$rYdSwxYw||S||ot j|S)zOCheck if a string is a URL or local file path with one of the given extensions.F)r'rpathlower ValueErrorendswithosisfile)rr)r,s r&_is_media_url_or_pathr2<st) D>>&,,..DD   55 }}Z((( ::<<  , , E1E1EEs&8 AAcP|drdSt|dS)z:Check if a string is an image URL, file path, or data URI.z data:image/T)z.jpgz.jpegz.pngz.gifz.bmpz.tiffz.webp)r$r2r%s r&is_image_url_or_pathr4Hs, }%%t 'b c ccr(ct|drdSt|sdS t|j}n#t$rYdSwxYw|dvS)z.Check if a string is a video URL or file path.)z.mp4z.aviz.movz.wmvz.flvz.mkvTF)zwww.youtube.comz youtube.comzyoutu.bez m.youtube.com)r2r'rnetlocr.)rr6s r&is_video_url_or_pathr7OsxT#STTt 4 u$& uu  s: AAc"t|dS)z/Check if a string is an audio URL or file path.)z.mp3z.wavz.oggz.flacz.aac)r2r%s r&is_audio_url_or_pathr9as 'P Q QQr(samplerct|ttfrt|dkrdSt|dtrt|dtrdS|D]W}t|t r d|vrd|vrdSt|tr |rt|dt rdSXdS)aCheck if a sample is a non-text pair (2-element tuple/list with at least one non-string element). Text pairs ``(str, str)`` are handled natively by tokenizers and detected as ``"text"`` modality by :func:`infer_modality`. This helper detects pairs that contain at least one non-string element (e.g. an image, audio array, or dict), which require conversion to message format. FrrolecontentT) isinstancetuplelistlenrdict)r:elems r&_is_non_text_pairrFfs fudm , ,F q0@0@u&)S!!jC&@&@u dD ! ! fnnd9J9J55 dD ! ! d z$q'4/H/H 55 4r( sampling_rateintextra_modality_kwargsdict[str, dict[str, Any]]Nonec|dd}|||krtd|d|d||dd<dS)a Record the batch-level ``sampling_rate``, raising if samples disagree. Feature extractors accept a single ``sampling_rate`` for the whole batch, so conflicting per-sample rates cannot be honored and would silently process some audio at the wrong rate. audiorGNz6This batch mixes audio with different sampling rates ( and z), but the processor accepts a single sampling rate per batch. Resample the audio to a common rate, or encode each sampling rate in a separate call.)getr.)rGrIexistings r&_record_sampling_raterQ{sz %W-11/BBHM 9 9 @U V V V7##J{L$I$I--//g13HIII|  Q ''--/// r( video_valuerc B|ddg}t|tr0||d|dSt t|t r|dt|}||jj |jj |jj tt|jjdd|jS|d|S)aUnwrap dict-wrapped video or a ``VideoDecoder`` into a raw array, collecting ``video_metadata``. Passes through unchanged if ``video_value`` is already a raw array/tensor/URL/path. Appends one metadata entry per video (``None`` when the sample carries none) so the batch-level list stays index-aligned with the videos. :func:`_reconcile_video_metadata` resolves the ``None`` entries once the whole batch has been parsed. videovideo_metadatarTNr)fpstotal_num_framesdurationframes_indices) setdefaultr@rDappendrOrget_frames_in_rangerCmetadata average_fps num_framesduration_secondsrBrangerXshape)r]rI metadata_list frame_batchs r& _unwrap_videorps*'2==>NPRSSM+t$$$[__-=>>???7##J{L$I$I!55a[9I9IJJ "+7$/$8$C'0A"&u[-=-CA-F'G'G"H"H       r( typed_inputs,list[tuple[Modality | Literal['pair'], Any]]c |d}|rd|vrdS|d}td|Dr |d=|s|d=dSd|D}tt||dD]r\}\}}| t |t st |dstd t|}|ddtt|d ||<sdS) aResolve the per-sample video metadata entries collected by :func:`_unwrap_video`. If no video in the batch carried metadata, the key is dropped so the processor infers defaults itself. Otherwise every ``None`` entry is filled with the frame count of that video, matching what transformers infers for a metadata-less video. This requires the video to be in memory, so path/URL videos without metadata raise instead, as passing a partial list would attach metadata to the wrong videos. r_r`Nc3K|]}|duV dSN).0entrys r& z,_reconcile_video_metadata..s& 4 4U5D= 4 4 4 4 4 4r(cg|]C\}}|dkst|t d|v$t|tr|dn|DS)r_r@rA)rwmodvalues r& z-_reconcile_video_metadata..sZ C '>>je44>C%S%00;ge9Gr(T)strict__len__zThis batch mixes videos that carry per-sample 'video_metadata' with videos that do not. Default metadata can only be inferred for in-memory videos (frame arrays or lists), so either pass 'video_metadata' for every video in the batch or for none of them.)rbrarcrd) rOall enumeratezipr@rhasattrr.rCrBrl) rqrI video_kwargsrnvideosindexrxr_rjs r&_reconcile_video_metadatarsQ),,W55L +<?? !12M 4 4m 4 4 444 ) * /%g.&F "+3}fT+R+R+R!S!S  ~u    eS ! !  )B)B a  ZZ *"5#4#455   e  r(ceZdZdZ d6d7d Zd8dZd9dZd:d;dZdd?d#Z d@d)Z e dAd,Z e d.,s(FFww("FFFFFFr(r)r!KNOWN_MODEL_TYPES_MESSAGE_FORMATSgetattrr@rany)rrstructured_patternsrs @r&rzInputFormatter._infer_formats ?? ? ?4T_E E9ot<<(C((  <    FFFF2EFFF F F <vr(inputslist[SingleInput | PairInput]Btuple[Modality, dict[str, list], defaultdict[str, dict[str, Any]]]c||sddgittfSg}tt}d}|D]F}t|r|d|fd},t |j}t |tr*|tvr!||hkr||}|dkrt||}n|dkrt||}nt |trJt|}d|vrt|d||d<d|vrt|d||d<n!|dkrt |tr|g}n|}|||fHt|||rg}|D]\} }| dkr)| |4| dkr>|r)| |x| dkr||t | tr|n| |i} || dd|i|fSt!|\} t#t%| } t'| d krW| }t |t*r|inYd } fd | Dn0t,d | d dfd|Did}||fS)a-Parse inputs and group by modality. Analyzes a list of inputs to detect their modality (text, image, audio, video, message) and groups them appropriately for the processor. Handles mixed modalities by converting to message format when necessary. Non-text pairs (e.g. ``(image, text)`` or ``(image, image)``) are detected and converted to message format with ``"query"``/``"document"`` roles via :meth:`pair_to_messages`. Args: inputs: List of inputs to parse. Can be: - str: Text inputs - tuple/list of str: Text pairs (for cross-encoders) - tuple/list of mixed types: Non-text pairs (e.g. image + text) - dict: Chat messages, audio data, or multimodal inputs - PIL.Image.Image: Image inputs - np.ndarray/torch.Tensor: Audio (1-2D) or video (3-5D) inputs Returns: Tuple of (modality, processor_inputs_dict, extra_modality_kwargs) where: - modality: Detected modality string (``"text"``, ``"image"``, etc.) or tuple of modalities - processor_inputs_dict: Dictionary mapping modality names to input lists - extra_modality_kwargs: Extra kwargs per modality (e.g. ``sampling_rate`` for audio) rFpairTrrMr_messager=rc0i|]fdDS)c g|] }| Srvrv)rwrxr|s r&r~z:InputFormatter.parse_inputs...s)S)S)S%*)S)S)Sr(rv)rwr|processed_inputss @r& z/InputFormatter.parse_inputs..s3#l#l#lX[C)S)S)S)SBR)S)S)S#l#l#lr(zMixed modalities detected: z!. Converting to 'message' format.crg|]3\}}t|tr|n||i4Srv) to_messager@rA)rwmodalityr}rs r&r~z/InputFormatter.parse_inputs..sR'%OOZ%-H-H$_EExY^N_``r()rrDrFrfinfer_modalityrr@r keysr\rprArpair_to_messages is_text_pairrrrBsetrCpoprloggerdebug)rrrqrI has_pairsitemrr}messagesr|typed modalitiesunique_modalities ordered_keysrs` @r& parse_inputszInputFormatter.parse_inputs1s8 ;FB<T):):: :EG +D 1 1 ! 3! 3D!&& ##VTN333 %dAZ[[[H$%% &(6J*J*Jtyy{{_g^hOhOhH~ 7""%d,ABBW$$%d,ABBHe,, T e##%25>CX%Y%YE'Ne##%25>CX%Y%YE'NY&&:dD+A+A&   5 1 2 2 2 2!,0EFFF  KH* < < U&==OOD$9$9%$@$@AAAAF]]t'8'8'?'?]OOD$9$9%$@$@AAAAI%%OOE****%/U%;%;MEE#uEOODOOE$:$:;;;;y(35JJ J'*L'9$ $ 011 OO  ! !Q & &(,,..H(C(( m$,.>#?   027799 #l#l#l#l_k#l#l#l  LLk7Hkkk l l l+7   !H)+@@@r(chat_template_kwargsdict[str, Any] | None str | Nonec6 tjdd}|dSpi|f}jjd|kr jdS fdtD^ } fdt t |D}|r&dd |D}d |d }n4d}n1#t$r$}d t|j d |}Yd}~nd}~wwxYwt d|pdtj ||f_|S)a~Why the chat template cannot carry a ``query``/``document`` pair, or None when it can. Decided by rendering probe pairs and diffing the outputs, never by reading the template source: templates can pass roles through without naming them (ChatML) or name them in branches a pair never reaches, and may transform content beyond recognition, so only a render that fails to move with its input proves the content goes nowhere. A template that raises is reported with the exception instead. The verdict is cached against a snapshot of the template and kwargs, so fixing either on a loaded model takes effect, which is the documented remedy. Args: chat_template_kwargs: The ``apply_chat_template`` kwargs the real render passes. These can select between named templates, so the probe must render with them. rNzthe model has no chat templaterr=c zg|]7}tjj|fddi8S)tokenizeF)rrapply_chat_templater)rwrrrs r&r~z5InputFormatter.pair_roles_failure..so  6DN6--d33>CG[r(c&g|] \}}|k |Srvrv)rwr>renderbases r&r~z5InputFormatter.pair_roles_failure..s#ZZZ f6UY>>t>>>r(rNc34K|]}t|VdSru)repr)rwr>s r&ryz4InputFormatter.pair_roles_failure..s($D$DDT$ZZ$D$D$D$D$D$Dr(zthe z+ content does not reach the rendered promptz it raised z: z'Pair-role probe for the chat template: zpair content survives)rrrPAIR_ROLE_PROBESr PAIR_ROLESjoin Exceptiontype__name__rrcopydeepcopy) rrrkeyvarieddroppedrolesfailureexcrs `` @r&pair_roles_failurez!InputFormatter.pair_roles_failures4>?DAA  3339r-.  ,1Fq1IS1P1P(+ +  - MD6[ZZZJ0G0GZZZG  $D$DG$D$D$DDDSSSS ? ? ?>499#5>>>>GGGGGG ?  cw?aJaccddd!%s!3!3W =s B.. C8CCmessages_batchlist[list[dict[str, Any]]]r!cbtttfd|DS)a*Whether any conversation in a batch carries both of the ``query``/``document`` pair roles. Anchored on produced messages so both routes into :meth:`pair_to_messages` are covered, and requiring both roles keeps hand-written single-role messages outside a check about pairs. c38K|]}d|DkVdS)c8h|]}|dS)r>rO)rwrs r& z:InputFormatter.has_pair_roles...s$!N!N!N''++f"5"5!N!N!Nr(Nrv)rwr pair_roless r&ryz0InputFormatter.has_pair_roles..s8nnS[:!N!NX!N!N!NNnnnnnnr()rrr)rrrs @r&has_pair_roleszInputFormatter.has_pair_roless4 __ nnnn_mnnnnnnr(r tuple | listlist[dict[str, Any]]ct\}}|\}}t|}t|}|jdkr ||d||dgSd}||||d||||dgS)aConvert a pair of inputs to query/document message format. Each element of the pair is wrapped in a message with role ``"query"`` (first element) or ``"document"`` (second element). The modality of each element is inferred individually via :func:`infer_modality`. Args: pair: A 2-element tuple or list of inputs (e.g. ``(image, text)``). Returns: List of two message dictionaries with ``"query"`` and ``"document"`` roles. rr>r?ct|tr#ttrfd|DSttr*|tvr!|hkr|d||igS)Nc0g|]}|vd|||iSrrv)rwr|rs r&r~zHInputFormatter.pair_to_messages.._to_content..s,YYY#STX[[c495[[[r(r)r@rArDr r)rrs `r& _to_contentz4InputFormatter.pair_to_messages.._to_contents(E** Zz$/E/E ZYYYYYYYY$%% &(6J*J*Jtyy{{_g^hOhOhH~Xx67 7r()rrr) rr query_roledoc_role query_itemdoc_itemquery_modality doc_modalityrs r&rzInputFormatter.pair_to_messagess * H# H' 33%h//  & ( (# ;;!h77   8 8 8 KK ,S,S T T++lH*M*M N N  r(user typed_inputdict[Modality, Any]r>c$|jdkrdt|dkr7tt|\}}||dgSt d|d|DdgS)aConvert a typed input dictionary to message format. Produces a single message with the given ``role``. For pair/multi-value inputs, use :meth:`pair_to_messages` instead (which is called automatically by :meth:`parse_inputs`). Args: typed_input: Dictionary mapping modality to input value (single value per modality). role: Role for the message (default: ``"user"``). Returns: List of message dictionaries (single message). rr=rzbFlat message format requested but multiple modalities detected. Falling back to structured format.c g|] \}}d|||i Srrv)rwrr}s r&r~z-InputFormatter.to_message..s%kkkOHeVXx?kkkr()rrCnextiteritemsrwarning)rrr>_r}s r&rzInputFormatter.to_messages  & ( (;1$$[%6%6%8%8 9 9::5!%%88999kkWbWhWhWjWjkkk    r(rr processor_inputsrD*tuple[Literal['message'], dict[str, list]]cv sddgifSt|tr|fn| ttt }g}t |D]Š fdD}t|dkrutt |\}}|dkr>||r)| | || | |dd|ifS)acConvert a batch of modality-specific inputs into the unified message format. Args: modality: The modality key (string) or tuple of modality keys. processor_inputs: Dictionary mapping modality names to lists of inputs. Returns: Tuple of ``("message", {"message": [messages_per_sample, ...]})`` rc6i|]}|v||Srvrv)rwr|irrs r&rz3InputFormatter.batch_to_message../s2jjjSX[_iXiXi3 0 5a 8XiXiXir(r=r) r@rrCrrvaluesrlrrrfrr) rrr batch_sizerrr|r}rrs ` @@r&batch_to_messagezInputFormatter.batch_to_messagesE  .y"o- -$.x$=$=Kh[[8 d#3#:#:#<#<==>>?? z"" : :AjjjjjjDTjjjK;1$$!${'8'8':':";";<< U&==T%6%6u%=%==OOD$9$9%$@$@AAA OODOOK88 9 9 9 99h///r(r}rct|ttfo+t|dkot d|DS)zVWhether a single input is a text pair, i.e. cross-encoder ``(query, document)`` input.r<c3@K|]}t|tVdSru)r@r)rwparts r&ryz.InputFormatter.is_text_pair..<s/KtKtfjJW[]`LaLaKtKtKtKtKtKtr()r@rArBrCr)r}s r&rzInputFormatter.is_text_pair9sD%%//tCJJ!OtKtKtnsKtKtKtHtHttr(c|D]h}|D]c}|d}t|tr-t|trt d|DrdS`dSidS)aCheck whether all messages in a batch contain only text content. Works with both flat format (``{"content": "hello"}``) and structured format (``{"content": [{"type": "text", "text": "hello"}]}``). Args: messages_batch: List of message lists, one per sample. Returns: True if every message contains only text, False if any contain non-text content. r?c3JK|]}|dddkVdS)rrNr)rwrs r&ryz7InputFormatter.is_text_only_messages..Qs5RR$488FF33v=RRRRRRr(FT)rOr@rrBr)rrrr?s r&is_text_only_messagesz$InputFormatter.is_text_only_messages>s' ! !H# ! !!++i00gs++gt,,!RR'RRRRR%$uuu%!555 !tr(rcg}|D]f}d|vsd|vrtd|d*|d}t|to|ot|dt}|jdkr|rt |dkr2d|dvr(|i|d|dditd t |d |||jd krO|sMt|tr!|i|dd|d gi:||Q||h|S) a;Normalize messages to the target format (``self.message_format``). Extra keys beyond ``"role"`` and ``"content"`` are preserved during conversion. Args: messages: List of message dictionaries to normalize. Returns: Normalized list of message dictionaries. r>r?zInvalid message format: z . Skipping.rrr=rz;Cannot convert structured message to flat format: contains z# content items. Keeping structured.rrr) rrr@rBrDrrCrfr)rr normalizedrr?is_currently_structureds r&normalize_messagesz!InputFormatter.normalize_messagesWs  + +GW$$ (@(@N'NNNOOOi(G&0$&?&?&lG&lPZ[bcd[egkPlPl #"f,,1H,w<<1$$71:)=)=%%&P&P)WQZ=O&P&PQQQQNNV$'LLVVV%%g....$ 44=T4gs++/%%&a&a)vW^>_>_=`&a&abbbb%%g....!!'****r(promptcR|jdkrfd|DSfd|DS)a'Prepend a system prompt to message format inputs. Args: messages: List of message lists (each message list represents one input). prompt: System prompt to prepend. Returns: Messages with system prompt prepended to each message list. rc"g|] }ddg|z S)systemrrvrw message_listrs r&r~z=InputFormatter.prepend_prompt_to_messages..s)hhh|h6::;lJhhhr(c*g|]}dddgdg|zS)rrr rrvrs r&r~z=InputFormatter.prepend_prompt_to_messages..sE   VV,L,L+M N N OR^ ^   r()r)rrrs `r&prepend_prompt_to_messagesz)InputFormatter.prepend_prompt_to_messages~sY  & ( (hhhh_ghhh h    (    r(texts'list[str | tuple[str, str] | list[str]]list[str | list[str]]c g}|D]g}t|tr|||z0|||dzgt|ddzh|S)abPrepend a prompt to text format inputs. For single texts, prepends the prompt directly. For text pairs (cross-encoder inputs), prepends only to the first text. Args: texts: List of text inputs (strings or pairs) prompt: Prompt to prepend Returns: Texts with prompt prepended rr=N)r@rrfrB)rrrresultrs r&prepend_prompt_to_textsz&InputFormatter.prepend_prompt_to_textss C CD$$$ C ftm,,,, vQ/04QRR>>ABBBB r()rNN)rrrr rrr rK)r r)rrr rru)rrr r)rrr r!)rrr r)r)rrr>rr r)rr rrDr r)r}rr r!)rrr r)rrrrr r)rrrrr r)r __module__ __qualname____doc__rrrrrrrr staticmethodrrr rrrvr(r&rrsi<)/6: 11111    DoAoAoAoAb*****Xoooo% % % % N     <0000:uuu\u\0%%%%N    &r(rSingleInput | PairInput | Anyrrr ctt|trdStt|trdStt|trdS|xtdx\t |r |d|vrdSdSxtdx\t |r |d|vrdSdSxtdx\t|r |d|vrdSdSxxtdx\nOx r"dkr\tdx\tdx\n*x r"dkr\tdx\tdx\nndSxtdx \d|vrd |vrd Sxtdx6\|r3t|d trd|d vrd |d vrd Sxtdx \d |vrd |vrdSxtdx \d |vrd|vrdSxtdx?\d |vr:tdt| txtdx\|rt| tz }|rg}d|vr|dd |vr|d|rdd|znd}tdt|tdttd|t#|dkrt%t'|St)t| Sxtdx\tdxt*jdx\nxt.jdx\nn<|jdvrdS|jdkrdS|jdvrdStd|jd td t5|jd!)"ahInfer the modality of a single input sample by inspecting its type/structure. Pure type-based detection, does not require a processor or tokenizer. Args: sample: A single input sample to inspect. supported_modalities: Optional list of modalities the model supports. When provided, string inputs that would be classified as image/video/audio based on URL/path heuristics are instead classified as ``"text"`` if that modality is not supported. This prevents misclassification of text that happens to contain media URLs. Returns: The detected modality string, or a tuple of modality strings for multimodal dict inputs with 2+ keys. A 1-key dict like ``{"image": pil}`` collapses to the bare modality string. Raises: ValueError: If the input type/structure is not recognized. NimagerMr_rvrr<r>r?rrrTrGr`zuDict input with 'array' key must also include 'sampling_rate' (for audio) or 'video_metadata' (for video). Got keys: raTo attach per-sample video metadata, pass the video itself as a dict, and keep processor options in processing_kwargs, e.g.: {"video": {"array": frames, "video_metadata": {"fps": 15, "total_num_frames": 3}}} processing_kwargs={"video": {"do_sample_frames": False}}zzTo attach the sampling rate, pass the audio itself as a dict, e.g.: {"audio": {"array": waveform, "sampling_rate": 16000}}r#z z;Multimodal dict input contains unrecognized modality keys: z. Expected keys from: .r=z-Empty dict input is not a valid input sample.)r=r<)z#Unsupported tensor dimensionality: z.-s(jjjX^.>RSSSjjjr(r=r)rCr)r[rrs ` r&infer_batch_modalityr8sQ$ vjjjjbijjjJ":!33:>>   Br(rcZt|trd|S|S)zSFormat a modality for display, e.g. ``("text", "image")`` becomes ``"text+image"``.+)r@rAr)rs r&format_modalityr;1s+(E"""xx!!! Or(rlist[Modality]sourcerc lddD}|dkr fd|D}n%#ttf$rt}YnwxYw|dhkrtd|d|dd|vrtd|d|d|rdd t |t D}d |D}t fd |D}|r/td |d|dd|d|d td |d|dt |trDtfd|Dr)td|dd|dtdt |d|d|d)aRaise a clear ``ValueError`` explaining why ``modality`` is not supported. Shared by :meth:`BaseModel.preprocess` and :meth:`Transformer.preprocess` so both raise the same, accurate guidance. ``source`` is a human-readable description of what rejected the input, e.g. ``"SentenceTransformer model"`` or ``"Transformer module"``. Args: inputs: The original batch of inputs, re-inspected per sample to distinguish explicit chat-style ``"message"`` inputs from inferred mixed-modality batches (both collapse to the ``"message"`` modality at the batch level). modality: The inferred (and unsupported) batch modality. supported_modalities: The modalities the source actually supports. source: Human-readable name of the rejecting component, used verbatim in the message. Raises: ValueError: Always, with a message tailored to the specific unsupported-modality scenario. z, c34K|]}t|VdSrur;rwms r&ryz3raise_unsupported_modality_error..Os*OOQoa00OOOOOOr(rc2h|]}t|Sr5r6r7s r&rz3raise_unsupported_modality_error..Vs4000V\v.js*#o#o1OA$6$6#o#o#o#o#o#or(r#cLh|]!}t|tr|n|fD]}|"Srvr{)rwrBrs r&rz3raise_unsupported_modality_error..ms?qqqQ[\]_dQeQeLoAAlmkoqq4tqqqqr(c3$K|] }|v|V dSrurvrwrrs r&ryz3raise_unsupported_modality_error..ns. f f$TQeEeEeEeEeEeEe f fr(z&This batch mixes multiple modalities (z ), but this z does not support z. Supported modalities: z), which this zZ cannot encode in a single batch. Encode each modality separately (one call per modality).c3 K|]}|vV dSrurvrGs r&ryz3raise_unsupported_modality_error..{s)*]*]D43G+G*]*]*]*]*]*]r(z supports rNzr individually, but cannot combine them in a single input. Encode each modality separately (one call per modality).z Modality 'z' is not supported by this ) rr. TypeErrorrr*r;r@rAr) rrrr= supported_strsample_modalities present_strbase_modalities unsupporteds ` r& raise_unsupported_modality_errorrO8s.IIOO:NOOOOOM9 &0000`f000  I& & & & #    &  + +QQQ@MQQQ  ) ) )tttcpttt   ))#o#oGX^m@n@n@n#o#o#oooKrq->qqqO f f f f/ f f fffK  d[ddV\dd#'99[#9#9ddS`dddfffTZfff (E"" s*]*]*]*]T\*]*]*]']']  hF h hgll8&<&< h h h    2_X.. 2 26 2 2!. 2 2 2  s8AA)rrr r!)rrr)r*r r!)r:rr r!)rGrHrIrJr rK)rRr rIrJr r)r]rrIrJr r)rqrrrIrJr rKru)r:r rrr r )r[rrrr r )rr r r) rrrr rr<r=rr r)7r __future__rrloggingr0 collectionsrtypingrrr urllib.parserrZr+r-)sentence_transformers.base.modality_typesr r r r r rr PIL.Imagerr) ImportErrortorchcodec.decodersrrOSError RuntimeError getLoggerrrrrrr'r2r4r7r9rFrQr\rprrrr8r;rOrvr(r&r\sCGG""""""  ######))))))))))!!!!!! +++++++HHH>>>>>>>>>Wl+LLLL  8 $ $  %%!# THHHH F F F Fdddd$RRRR * D D D D "6) ) ) ) XCCCCCCCCP37mmmmmd37CCCCC0LLLLLLs$A AAA$$ A32A3