+ j:R^RIt^RIt^RIt^RIHt^RIHt^RI H t H t ^RI H t HtHtHt^RIHtHt^RIHt^RIHtHtHt^R IHt^R IHt^R IHt^R I H!t!!R R]4t"!RR]RR7t#!RR]4t$!RR]4t%!RR]RR7t&!RR]!4t'.ROt(R#)N) functional) BatchFeature)group_images_by_shapereorder_images)ChannelDimensionPILImageResamplingSizeDictget_image_size) ImagesKwargs VideosKwargs) TensorType) VideoMetadatagroup_videos_by_shapereorder_videos) Glm46VConfig)Glm46VImageProcessor)Glm46VImageProcessorPil)Glm46VVideoProcessorc]tRt^)tRtRtRtR#) GlmgaConfigar image_start_token_id (`int`, *optional*, defaults to 151339): The image start token index to encode the start of image. image_end_token_id (`int`, *optional*, defaults to 151340): The image end token index to encode the end of image. video_start_token_id (`int`, *optional*, defaults to 151361): The video start token index to encode the start of video. video_end_token_id (`int`, *optional*, defaults to 151362): The video end token index to encode the end of video. ```python >>> from transformers import AutoModelForImageTextToText, GlmgaConfig >>> # Initializing a Glmga style configuration >>> configuration = GlmgaConfig() >>> # Initializing a model (reusing the GLM-4.6V implementation) from that configuration >>> model = AutoModelForImageTextToText.from_config(configuration) >>> # Accessing the model configuration >>> configuration = model.config ```glmgaN)__name__ __module__ __qualname____firstlineno____doc__ model_type__static_attributes__rh/home/luca/screen-checking/.venv/lib/python3.14/site-packages/transformers/models/glmga/modular_glmga.pyrr)s .Jr rc0a]tRt^DtoRtV3RltRtVtR#)GlmgaImageProcessorKwargsa patch_size (`int`, *optional*, defaults to 14): The spatial patch size of the vision encoder. temporal_patch_size (`int`, *optional*, defaults to 2): The temporal patch size of the vision encoder. merge_size (`int`, *optional*, defaults to 2): The merge size of the vision encoder to llm encoder. patch_expand_factor (`int`, *optional*, defaults to 1): The patch_expand_factor of the vision encoder to llm encoder. cJ<V^8dQh/S[;R&S[;R&S[;R&S[;R&#) patch_sizetemporal_patch_size merge_sizepatch_expand_factor)int)format __classdict__s"r! __annotate__&GlmgaImageProcessorKwargs.__annotate__Ds7OOr rNrrrrr__annotate_func__r__classdictcell__r,s@r!r#r#Ds r r#F)totalc6a]tRt^Vto^tV3RlRltRtVtR#)GlmgaImageProcessorc <V^8dQhRS[R,RS[RS[RRRS[RS[R S[R S[S[S[,,R ,R S[S[S[,,R ,R S[RS[RS[RS[RS[R ,RS[S[,R ,RS[/#)r%imagesz torch.Tensor do_resizesizeresample7PILImageResampling | tvF.InterpolationMode | int | None do_rescalerescale_factor do_normalize image_meanN image_stdr&r'r(r)disable_groupingreturn_tensorsreturn)listboolr floatr*strr r)r+r,s"r!r- GlmgaImageProcessor.__annotate__Ys? ? ^$? ?  ? L ?  ? ? ? DK'$.? 4;&-? ? !? ? !? +? j(4/!? $ %? r c \WR7wpp/pVP4F6wppV'd$VPVVVW,V ,V R7pVVV&K8 \VV4p\VVR7wpp/p/pVP4FQwppVP VWVWxV 4pVP VV V V R7wpppVVV&^VV..\ V4,VV&KS \VV4p\VV4p\ V4^8Xd V^,M\P!V^R7p\P!V4p\RVRV/VR7#)z) Preprocess an image or batch of images. )rA)r7r9r:factortemporal_factorr&r(r'dim pixel_valuesimage_grid_thwdata tensor_type) ritemsresizerrescale_and_normalizepatchifylentorchcattensorr)selfr7r8r9r:r<r=r>r?r@r&r'r(r)rArBkwargsgrouped_imagesgrouped_images_indexresized_images_groupedshapestacked_imagesresized_imagesprocessed_images_groupedprocessed_gridspatchesgrid_hgrid_wprocessed_imagesrOrPs&&&&&&&&&&&&&&&&, r! _preprocessGlmgaImageProcessor._preprocessYs.0EV/o,,!#%3%9%9%; !E>!%)%%25HH$7 "-"-; "5 )&<((>@TU/D^fv/w,,#% %3%9%9%; !E>!77 LV_N'+mm%%$7 '4' #GVV/6 $U +'(&&&9%:S=P%POE "&<**BDXY(:NO.12B.Cq.H'*eiiXhnoNp o6 ,0@.Q_m  r rNrrrrr)rjrr1r2s@r!r5r5Vs? ? r r5c6a]tRt^to^tV3RlRltRtVtR#)GlmgaImageProcessorPilc<V^8dQhRS[S[P,RS[RS[RRRS[RS[RS[R S[S[S[,,R ,R S[S[S[,,R ,R S[R S[RS[RS[RS[S[,R ,RS[ /#)r%r7r8r9r:zPILImageResampling | int | Noner<r=r>r?Nr@r)r&r'r(rBrC) rDnpndarrayrEr rFr*rGr r)r+r,s"r!r-#GlmgaImageProcessorPil.__annotate__s9 9 RZZ 9 9  9 4 9  9 9 9 DK'$.9 4;&-9 !9 9 !9 9 j(4/9 " #9 r c .p.pVFpV'dVPVVVW,V R7pV'dVPVV4pV'dVPVW4pVPVV V V R7wpppVP V4VP ^VV.4K \ P !V^R7p\ P!V4p\RVRV/VR7#)z/ Preprocess images one by one for PIL backend. )r9r:rJrKrL)axisrOrPrQ) rUrescale normalizerWappendrp concatenatearrayr)r\r7r8r9r:r<r=r>r?r@r)r&r'r(rBr]rireimagerfrgrhrOrPs&&&&&&&&&&&&&&&, r!rj"GlmgaImageProcessorPil._preprocesss*E %%2$7 $ UN;ujD&*mm%%$7 '4' #GVV  # #G ,  " "Avv#6 738~~&6Q? /2 ,0@.Q_m  r rNrlr2s@r!rnrns9 9 r rnc0a]tRt^toRtV3RltRtVtR#)GlmgaVideoProcessorInitKwargsa patch_size (`int`, *optional*, defaults to 14): The spatial patch size of the vision encoder. temporal_patch_size (`int`, *optional*, defaults to 2): The temporal patch size of the vision encoder. merge_size (`int`, *optional*, defaults to 2): The merge size of the vision encoder to llm encoder. max_frames (`int`, *optional*, defaults to 640): The maximum number of frames that can be sampled. max_image_size (`dict`, *optional*, defaults to `28 * 28 * 2 * 55790`): The maximum pixels a video can be resized to. patch_expand_factor (`int`, *optional*, defaults to 1): The factor that will be multiplied to `patch_size` when resizing videos. cx<V^8dQh/S[S[S[3,;R&S[;R&S[;R&S[;R&S[;R&S[;R&#)r%max_image_sizer&r'r(r) max_frames)dictrGr*)r+r,s"r!r-*GlmgaVideoProcessorInitKwargs.__annotate__sX cN"!"O#$%&O'()*O+r rNr/r2s@r!r}r}s r r}ca]tRt^toRR RR /tRR /t^t^tRtR V3RlRllt RRR] PRRRRRRRRRR3V3RlR llt R t VtR#)GlmgaVideoProcessor shortest_edge longest_edgeiNcD<V^8dQhRS[RS[S[,R,/#)r%metadatafpsN)rr*rF)r+r,s"r!r- GlmgaVideoProcessor.__annotate__s,A)A)A)5[4 A)r c xVe\VRR4f \R4hVPpV^, pVP;'g#\ WQP , 4^,pVeTM VP p\ Wg,4p\WP4p^VP , p \V4U u.uF qV ,NK p p WH8d=\V4U u.uF&p \P!W,V, 4NK( p p MW.p ^p^V, p\V4F;pV V,V8gKW, pV PV4WV, 8gK;M \V 4V8df\V 4^8Xd^\V^, ^4ppMV ^,V R,pp\P !VVV\ R7P#4p MC\V 4V8d4\P !^V^, V\ R7P#4p \%4.ppV F.pVV9gK VP'V4VPV4K0 \V4^,'dVPVR,4\P(!V\ R7#uup iuup i)a? Args: metadata (`VideoMetadata`): Metadata of the video containing information about total duration, fps and total number of frames. fps (`int` or `float`, *optional*): Target frames to sample per second. Defaults to `self.fps`. Returns: np.ndarray: Indices to sample video frames. NrzAsked to sample frames per second but no video metadata was provided which is required when sampling in Glmga. Please pass in `VideoMetadata` object or set `do_sample_frames=False`)dtype)getattr ValueErrortotal_num_framesdurationroundrr*minrrangemathfloorrwrXmaxrplinspacetolistsetaddry)r\rrr] total_frames max_frame_idxr target_fps extract_tduration_per_framei timestamps_i frame_indicescurrent_secondinv_fps frame_indexstartendseenuniqidxs&&&, r! sample_frames!GlmgaVideoProcessor.sample_framessK  wx=EX   00 $q( $$OOmll.J(Ka(OOS -.  ??3 -6;L6IJ6I,,,6I J  #QVW`QabQa2TZZ(9I(EFQaMbMMN*nG$\2 k*n<"-N!((5%G);; 3 }  )=!Q&L1$4a 8ss*1-}R/@sKKsISIPPRM  ) +KKr?r@r)r&r'r(rB) rDrYTensorrEr rFr*rGr )r+r,s"r!r-r<sCCCCU\\"CCCC CC o CC L CCCCCCCCDK'$.CC4;&-CC!4ZCC$JCC!4ZCC$JCC j(4/!CCr c \V4wpp/pVP4FOwppV'dVPV4pV'd$VPVVVW,V ,V R7pVVV&KQ \ VV4p\V4wpp/p/pVP4Fwwpp\ V^,\ PR7wppVPVWgWV 4pVPVV VV R7wppppVVV&VVV..\V4,VV&Ky \ VV4p\ VV4p\P!V^R7p \P!V4p!RV RV!/p"\V"VR7#))rr9r:rJrK) channel_dimrLrMpixel_values_videosvideo_grid_thwrQ)rrTconvert_to_rgbrUrr rFIRSTrVrWrXrYrZr[r)#r\rrr8r9r:r<r=r>r?r@r)r&r'r(rBr]grouped_videosgrouped_videos_indexresized_videos_groupedrastacked_videosresized_videosprocessed_videos_groupedreresized_height resized_widthrfgrid_trgrhprocessed_videosrrrRs#&&&&&&&&&&&&&&&&, r!rjGlmgaVideoProcessor._preprocess<s&0EV/L,,!#%3%9%9%; !E>!%!4!4^!D!%)%%25HH$7 "-"-; "5 )&<((>@TU0E^/T,,#% %3%9%9%; !E>,:>!;LZjZpZp,q )NM"77 LV_N/3mm%%$7 /</ +GVVV/6 $U +'-vv&>%?#nBU%UOE "&<"**BDXY(:NO#ii(8a@o6 !#6 n  >BBr ri1i6)Ngp?)rrrrr9rrr)rrrBICUBICrjrr1r2s@r!rrs Y8K LD$&9:N CJA)A)L $ $N`NhNh )!1504*.!%*.!%26!CCCCCCr r)rr5rnr))rnumpyrprYtorchvision.transforms.v2rtvFimage_processing_utilsrimage_transformsrr image_utilsrrr r processing_utilsr r utilsr video_utilsrrrglm46v.configuration_glm46vrglm46v.image_processing_glm46vr"glm46v.image_processing_pil_glm46vrglm46v.video_processing_glm46vrrr#r5rnr}r__all__rr r!rs  72E ;OO6AHA ,6 E$B .B J< 4< ~L0MC.MC` r