/j16ddlmZddlmZmZddlmZddlZddlmZm Z ddl m Z ddl m Z mZddlmZdd lmZmZmZdd lmZd d lmZd dlmZGdde ZGdde ZGddeeZGddeeZGddeZdS)) annotations)copydeepcopy)PathN)YOLOConcatDatasetbuild_yolo_dataset)LoadVisualPrompt)DetectionTrainerDetectionValidator) YOLOEModel) DEFAULT_CFGLOGGERRANK) unwrap_model)WorldTrainerFromScratch)YOLOEDetectValidatorcFeZdZdZeddfdfd Zddd Zd ZdddZxZ S) YOLOETrainera`A trainer class for YOLOE object detection models. This class extends DetectionTrainer to provide specialized training functionality for YOLOE models, including custom model initialization, validation, and dataset building with multi-modal support. Attributes: loss_names (tuple): Names of loss components used during training. Methods: get_model: Initialize and return a YOLOEModel with specified configuration. get_validator: Return a YOLOEDetectValidator for model validation. build_dataset: Build YOLO dataset with multi-modal support for training. N overrides dict | None _callbacksc|i}|drJd|ddd|d<t|||dS)aInitialize the YOLOE Trainer with specified configurations. Args: cfg (dict): Configuration dictionary with default training settings from DEFAULT_CFG. overrides (dict, optional): Dictionary of parameter overrides for the default configuration. _callbacks (dict, optional): Dictionary of callback functions to be applied during training. NcompilezTraining with 'model=modelz' requires 'compile=False'F overlap_mask)getsuper__init__)selfcfgrr __class__s h/home/longshao/multi-rider-rag/.venv/lib/python3.11/site-packages/ultralytics/models/yolo/yoloe/train.pyr zYOLOETrainer.__init__$sm  I==++ss-sYwEW-s-s-sss+$) .! i44444Tverboseboolctt|tr|dn||jdt |jdd|o t dk}|r|||S)aOReturn a YOLOEModel initialized with the specified configuration and weights. Args: cfg (dict | str, optional): Model configuration. Can be a dictionary containing a 'yaml_file' key, a direct path to a YAML file, or None to use default configuration. weights (str | Path, optional): Path to pretrained weights file to load into the model. verbose (bool): Whether to display model information during initialization. Returns: (YOLOEModel): The initialized YOLOE model. Notes: - The number of classes (nc) is hard-coded to a maximum of 80 following the official configuration. - The nc parameter here represents the maximum number of different text samples in one image, rather than the actual number of classes. yaml_filechannelsncPchr+r&)r isinstancedictdataminrload)r!r"weightsr&rs r$ get_modelzYOLOETrainer.get_model2s~& *3 5 5 >C  3y$49T?B''*      JJw    r%cxd|_t|j|jt |j|jS)z9Return a YOLOEDetectValidator for YOLOE model validation.boxclsdflsave_dirargsr) loss_namesr test_loaderr=rr> callbacksr!s r$ get_validatorzYOLOETrainer.get_validatorPs:-#  t}4 ??W[We    r%trainimg_pathstrmodebatch int | Nonec tt|jr+t|jjndd}t |j|||j||dk||dkS)aBuild YOLO Dataset. Args: img_path (str): Path to the folder containing images. mode (str): 'train' mode or 'val' mode, users are able to customize different augmentations for each mode. batch (int, optional): Size of batches, this is for rectangular training. Returns: (Dataset): YOLO dataset configured for training or validation. r valrD)rGrectstride multi_modal)maxintrrrNrr>r2)r!rErGrHgss r$ build_datasetzYOLOETrainer.build_datasetWsydjO\$*--488:::aPPRT U U! Ix 45=Y[imqxix    r%)rrrrNNTr&r'rDN)rErFrGrFrHrI) __name__ __module__ __qualname____doc__r r r6rCrS __classcell__r#s@r$rrs  'ae 5 5 5 5 5 5 5<            r%rceZdZdZdddZdS) YOLOEPETraineraHFine-tune YOLOE model using linear probing approach. This trainer freezes most model layers and only trains specific projection layers for efficient fine-tuning on new datasets while preserving pretrained features. Methods: get_model: Initialize YOLOEModel with frozen layers except projection layers. NTr&r'ctt|tr|dn||jd|jd|o tdk}|jd`| Jd|r|||t|jd }| |}| |||jd |jt|jdjd d d |jdjd d <t|jdjd d d |jdjd d <t|jdjd d d |jdjd d <t%|jdd dt|jdjd d d |jdjd d <t|jdjd d d |jdjd d <t|jdjd d d |jdjd d <||S)aReturn YOLOEModel initialized with specified config and weights. Args: cfg (dict | str, optional): Model configuration. weights (str, optional): Path to pretrained weights. verbose (bool): Whether to display model information. Returns: (YOLOEModel): Initialized model with frozen layers except for specific projection layers. r)r*r+r-r.Nz7Pretrained weights must be provided for linear probing.namesrrTr one2one_cv3)r r0r1r2rrsavper4evallistvalues get_text_pe set_classesfusepercv3requires_grad_getattrrarD)r!r"r5r&rr`tpes r$r6zYOLOEPETrainer.get_modelrs *3 5 5 >C  3y$y*     KO !""$]"""  JJw    TYw'..0011&& %%%% BUX&&&$,U[_-@-CA-F$G$G$V$VW[$\$\ BAq!$,U[_-@-CA-F$G$G$V$VW[$\$\ BAq!$,U[_-@-CA-F$G$G$V$VW[$\$\ BAq! 5;r?M4 8 8 D08R9LQ9OPQ9R0S0S0b0bcg0h0hEKO ' *1 -08R9LQ9OPQ9R0S0S0b0bcg0h0hEKO ' *1 -08R9LQ9OPQ9R0S0S0b0bcg0h0hEKO ' *1 -  r%rTrU)rWrXrYrZr6r%r$r^r^hs7,,,,,,,r%r^c$eZdZdZddd ZddZdS)YOLOETrainerFromScratchaTrain YOLOE models from scratch with text embedding support. This trainer combines YOLOE training capabilities with world training features, enabling training from scratch with text embeddings and grounding datasets. Methods: build_dataset: Build datasets for training with grounding support. generate_text_embeddings: Generate and cache text embeddings for training. rDNrElist[str] | strrGrFrHrIc0tj||||S)aBuild YOLO Dataset for training or validation. This method constructs appropriate datasets based on the mode and input paths, handling both standard YOLO datasets and grounding datasets with different formats. Args: img_path (list[str] | str): Path to the folder containing images or list of paths. mode (str): 'train' mode or 'val' mode, allowing customized augmentations for each mode. batch (int, optional): Size of batches, used for rectangular training/validation. Returns: (YOLOConcatDataset | Dataset): The constructed dataset for training or validation. )rrS)r!rErGrHs r$rSz%YOLOETrainerFromScratch.build_datasets'4T8T5QQQr%texts list[str]rQ cache_dirrct|jj}|d|dddddz }|rgt jd|dtj||j }t| t|kr|St jd |dt|j ||d d }tt||d }tj|||S)abGenerate text embeddings for a list of text samples. Args: texts (list[str]): List of text samples to encode. batch (int): Batch size for processing. cache_dir (Path): Directory to save/load cached embeddings. Returns: (dict): Dictionary mapping text samples to their embeddings. text_embeddings_:_/z.ptzReading existed cache from '') map_locationzCaching text embeddings to 'TF)without_reprtacache_clip_modelr)rr text_modelreplaceexistsrinfotorchr4devicesortedkeysrfr1zipsqueezesave)r!rsrHrur cache_pathtxt_map txt_featss r$generate_text_embeddingsz0YOLOETrainerFromScratch.generate_text_embeddingss8TZ((3!bEMM#s4K4K4S4STWY\4]4]!b!b!bb       KDzDDD E E Ej$+FFFGgllnn%%66 @:@@@AAA ,,88VZmr8ss s5)"3"3A"6"67788 7J'''r%rVrErqrGrFrHrI)rsrtrHrQrur)rWrXrYrZrSrrnr%r$rprpsPRRRRR r%rpc&eZdZdZdZdZddZdS) YOLOEPEFreeTraineraTrain prompt-free YOLOE model. This trainer combines linear probing capabilities with from-scratch training for prompt-free YOLOE models that don't require text prompts during inference. Methods: get_validator: Return standard DetectionValidator for validation. preprocess_batch: Preprocess batches without text features. set_text_embeddings: Set text embeddings for datasets (no-op for prompt-free). cxd|_t|j|jt |j|jS)z6Return a DetectionValidator for YOLO model validation.r8r<)r?r r@r=rr>rArBs r$rCz YOLOEPEFreeTrainer.get_validators:-!  t}4 ??W[We    r%c,tj||S)z_Preprocess a batch of images for YOLOE training, adjusting formatting and dimensions as needed.)r preprocess_batch)r!rHs r$rz#YOLOEPEFreeTrainer.preprocess_batchs0u===r%rHrQcdS)aNo-op override for prompt-free training that does not require text embeddings. Args: datasets (list[Dataset]): List of datasets containing category names to process. batch (int): Batch size for processing text embeddings. Nrn)r!datasetsrHs r$set_text_embeddingsz&YOLOEPEFreeTrainer.set_text_embeddingss  r%N)rHrQ)rWrXrYrZrCrrrnr%r$rrsP     >>>      r%rc0eZdZdZd d fd Zfd ZxZS)YOLOEVPTrainera4Train YOLOE model with visual prompts. This trainer extends YOLOETrainerFromScratch to support visual prompt-based training, where visual cues are provided alongside images to guide the detection process. Methods: build_dataset: Build dataset with visual prompt loading transforms. rDNrErqrGrFrHrIc&t|||}t|tr1|jD](}|jt)n&|jt|S)a$Build YOLO Dataset for training or validation with visual prompts. Args: img_path (list[str] | str): Path to the folder containing images or list of paths. mode (str): 'train' mode or 'val' mode, allowing customized augmentations for each mode. batch (int, optional): Size of batches, used for rectangular training/validation. Returns: (YOLOConcatDataset | Dataset): YOLO dataset configured for training or validation, with visual prompts for training mode. )rrSr0rr transformsappendr )r!rErGrHdatasetdr#s r$rSzYOLOEVPTrainer.build_datasets''''$>> g0 1 1 :% 8 8 ##$4$6$67777 8   % %&6&8&8 9 9 9r%c^tt|jjt r<|jjjD](}|jt)dS|jjjtdS)zPClose mosaic augmentation and add visual prompt loading to the training dataset.N) r_close_dataloader_mosaicr0 train_loaderrrrrrr )r!rr#s r$rz'YOLOEVPTrainer._close_dataloader_mosaics ((*** d'/1B C C L&.7 8 8 ##$4$6$67777 8 8   % 0 7 78H8J8J K K K K Kr%rVr)rWrXrYrZrSrr[r\s@r$rrsn(LLLLLLLLLr%r) __future__rrrpathlibrrultralytics.datarrultralytics.data.augmentr ultralytics.models.yolo.detectr r ultralytics.nn.tasksr ultralytics.utilsr rrultralytics.utils.torch_utilsrworld.train_worldrrLrrr^rprrrnr%r$rs#""""" BBBBBBBB555555OOOOOOOO++++++7777777777666666777777%%%%%%P P P P P #P P P f66666%666r11111l,C111h     )@   B%L%L%L%L%L,%L%L%L%L%Lr%