o پi @s@ddlmZmZddlmZddlmZmZGdddeZdS))ListUnion)Sarashina2VisionForCausalLM)BaseMultimodalProcessorMultimodalSpecialTokenscs<eZdZegZfddZdeeee ffddZ Z S)Sarashina2VisionProcessorcstj|||g|Ri|d|_t|dd|_t|dd|_t|dd|_t|j|jd||_ t |d rNt |j d rP|j j fd d }||j _ dSdSdS) Nz<|file|>image_token_indexstart_image_token_indexiend_image_token_indexi) image_tokenimage_token_idimage_processor _preprocesscs,hdfdd|D}|i|S)N> resample do_resize image_std do_rescale image_mean data_format do_normalizedo_convert_rgbrescale_factorinput_data_formatcsi|] \}}|vr||qSr).0kvallowed_paramsrf/home/ubuntu/.local/lib/python3.10/site-packages/sglang/srt/multimodal/processors/sarashina2_vision.py 0szRSarashina2VisionProcessor.__init__..patched_preprocess..)items)argskwargsfiltered_kwargsoriginal_preprocessrr patched_preprocess!s  z>Sarashina2VisionProcessor.__init__..patched_preprocess) super__init__ IMAGE_TOKENgetattr IM_TOKEN_ID IM_START_ID IM_END_IDrbuild mm_tokenshasattrrr)self hf_config server_args _processorr#r$r( __class__r&r r* s$   z"Sarashina2VisionProcessor.__init__ image_datac sF|j|||jd}|j||jd\}}} |||jj|j|jdS)zLProcess image data for Sarashina2Vision model using standard SGLang pattern.)promptr9multimodal_tokens) base_outputr1)mm_items input_ids im_token_id im_start_id im_end_id) load_mm_datar1process_and_combine_mm_datatolistr r.r/) r3r9 input_text request_objr#r$r<r=r>retrrr process_mm_data_async7s   z/Sarashina2VisionProcessor.process_mm_data_async) __name__ __module__ __qualname__rmodelsr*rrstrbytesrH __classcell__rrr7r r s  *rN) typingrr#sglang.srt.models.sarashina2_visionr/sglang.srt.multimodal.processors.base_processorrrrrrrr s