o *jv @szddlZddlZddlZddlZddlmZmZmZddlZddl Zddl Z ddl m Z ddl mZmZmZmZmZddlmZddlmZddlmZz dd l mZejZWn eyce jZYnweje jejd krved d d gZ ddZ!ddZ"e j#$rdndddfde%dee%e j&fde'de%fdd Z(  d dee%ee%fde)de'dee j*e j+ffdd Z,dS)!N)AnyListUnion)Image) CenterCropCompose NormalizeResizeToTensor)tqdm) build_model)SimpleTokenizer)InterpolationModez1.7.1z.PyTorch version 1.7.1 or higher is recommendedloadtokenizecCs |dS)NRGB)convert)imagers/var/www/html/Deteccion_Ine/venv/lib/python3.10/site-packages/modelscope/models/cv/text_driven_segmentation/clip.py_convert_image_to_rgb"s rcCs&tt|tdt|tttddgS)N) interpolation)g3<4'?gwg M?gy{ ?)gB91?gwt.?g U?)rr BICUBICrrr r)Zn_pxrrr _transform&s rcudacpuFnamedevicejitrootc s|st}tdkr||t|jjfStjj fddgd}dd|j dDdfd d }| |||j ||jtdkrtjj d dgd}t|j d d }|fdd} | | | |j | |j||t|jfS)NrcstgtSN)torchonestorr)rrr?szload..)Zexample_inputscSsg|] }dt|vr|qS)ZDevice)repr).0nrrr @s  zload..prim::Constantcsz t|dr |jgng}Wn tyg}Ynwt|dr%||jj|D]}|dD]}d|vrDt|ddrD| q.q'dS)Ngraphforward1r*valuer) hasattrr, RuntimeErrorappendr- findAllNodesZattributeNamesstr startswithcopyAttributes)modulegraphsr,node) device_noderr patch_deviceEs$   zload..patch_devicecSstgSr!)r"r#floatrrrrr%[saten::tor csz t|dr |jgng}Wn tyg}Ynwt|dr%||jj|D](}|dD] }t|}dD]}||ddkrM|| q8q.q'dS)Nr,r-r<)r r.) r/r,r0r1r-r2listinputsr8r5)r6r7r,r8r@i) float_noderr patch_float_s"   zload..patch_float)r r$r3r;rZvisualZinput_resolutionr"rtracer,r2applyZ encode_imageZ encode_textr?ZfindNoder@r8item) rrrr modelZ device_holderr:Z float_holderZ float_inputrCr)rr9rBrr1s:             Mtextscontext_lengthtruncatereturncst|tr|g}jdjdfdd|D}tjtjtjdkr6tjt ||tj d}n tjt ||tj d}t |D]0\}}t ||krh|r\|d|}|d<n t d ||d |t|||dt |f<qE|S) a Returns the tokenized representation of given input string(s) Parameters ---------- texts : Union[str, List[str]] An input string or a list of input strings to tokenize context_length : int The context length to use; all CLIP models use 77 as the context length truncate: bool Whether to truncate the text in case its encoding is longer than the context length Returns ------- A two-dimensional tensor containing the resulting tokens, shape = [number of input strings, context_length]. We return LongTensor when torch version is <1.8.0, since older index_select requires indices to be long. z<|startoftext|>z <|endoftext|>cs"g|] }g|gqSr)encode)r'text _tokenizerZ eot_tokenZ sot_tokenrrr)sztokenize..z1.8.0)ZdtypeNr+zInput z is too long for context length ) isinstancer3encoder packagingversionparser" __version__Zzeroslenlongint enumerater0Ztensor)rPrIrJrKZ all_tokensresultrAtokensrrOrrzs0       )rHF)-hashlibosurllibwarningstypingrrrrSZpackaging.versionr"ZPILrZtorchvision.transformsrrrr r r rGr Zsimple_tokenizerrZ _Tokenizerrr ImportErrorrTrUrVwarn__all__rrrZ is_availabler3rboolrrYZ IntTensorZ LongTensorrrrrrsf             L