.\YOLO-World\configs\deploy\detection_onnxruntime-fp16_dynamic.py
# 设置基础路径为指定的配置文件路径 _base_ = ( '../../third_party/mmdeploy/configs/mmdet/detection/' 'detection_onnxruntime-fp16_dynamic.py') # 定义代码库配置，包括模型类型、任务类型、后处理参数等 codebase_config = dict( type='mmyolo', task='ObjectDetection', model_type='end2end', post_processing=dict( score_threshold=0.1, confidence_threshold=0.005, iou_threshold=0.3, max_output_boxes_per_class=100, pre_top_k=1000, keep_top_k=100, background_label_id=-1), module=['mmyolo.deploy']) # 定义后端配置，使用onnxruntime作为后端 backend_config = dict( type='onnxruntime')
.\YOLO-World\configs\deploy\detection_onnxruntime-int8_dynamic.py
# 设置基础路径为指定的配置文件路径 _base_ = ( '../../third_party/mmdeploy/configs/mmdet/detection/' 'detection_onnxruntime-fp16_dynamic.py') # 定义后端配置，设置精度为int8 backend_config = dict( precision='int8') # 定义代码库配置，包括模型类型、任务类型、后处理参数等 codebase_config = dict( type='mmyolo', task='ObjectDetection', model_type='end2end', post_processing=dict( score_threshold=0.1, confidence_threshold=0.005, iou_threshold=0.3, max_output_boxes_per_class=100, pre_top_k=1000, keep_top_k=100, background_label_id=-1), module=['mmyolo.deploy']) # 重新定义后端配置，设置类型为onnxruntime backend_config = dict( type='onnxruntime')
.\YOLO-World\configs\deploy\detection_onnxruntime_static.py
# 设置基础路径为指定的配置文件路径 _base_ = ( '../../third_party/mmyolo/configs/deploy/' 'detection_onnxruntime_static.py') # 定义代码库配置，包括类型、任务、模型类型和后处理参数 codebase_config = dict( type='mmyolo', task='ObjectDetection', model_type='end2end', post_processing=dict( score_threshold=0.25, confidence_threshold=0.005, iou_threshold=0.65, max_output_boxes_per_class=200, pre_top_k=1000, keep_top_k=100, background_label_id=-1), module=['mmyolo.deploy']) # 定义后端配置，类型为onnxruntime backend_config = dict( type='onnxruntime')
.\YOLO-World\configs\deploy\detection_tensorrt-fp16_static-640x640.py
_base_ = ( '../../third_party/mmyolo/configs/deploy/' 'detection_tensorrt-fp16_static-640x640.py') # 设置基础配置文件路径 onnx_config = dict( type='onnx', export_params=True, keep_initializers_as_inputs=False, opset_version=11, save_file='end2end.onnx', input_names=['input'], output_names=['dets', 'labels'], input_shape=(640, 640), optimize=True) # 设置 ONNX 配置参数，包括类型、是否导出参数、是否保留初始化器作为输入、操作集版本、保存文件名、输入输出名称、输入形状、是否优化 backend_config = dict( type='tensorrt', common_config=dict(fp16_mode=True, max_workspace_size=1 << 34), model_inputs=[ dict( input_shapes=dict( input=dict( min_shape=[1, 3, 640, 640], opt_shape=[1, 3, 640, 640], max_shape=[1, 3, 640, 640]))) ]) # 设置后端配置参数，包括类型、通用配置、模型输入 use_efficientnms = False # whether to replace TRTBatchedNMS plugin with EfficientNMS plugin # noqa E501 # 是否使用 EfficientNMS 插件替换 TRTBatchedNMS 插件 codebase_config = dict( type='mmyolo', task='ObjectDetection', model_type='end2end', post_processing=dict( score_threshold=0.25, confidence_threshold=0.005, iou_threshold=0.65, max_output_boxes_per_class=100, pre_top_k=1, keep_top_k=1, background_label_id=-1), module=['mmyolo.deploy']) # 设置代码库配置参数，包括类型、任务、模型类型、后处理参数、模块
.\YOLO-World\configs\deploy\detection_tensorrt-int8_static-640x640.py
_base_ = [ '../../third_party/mmdeploy/configs/mmdet/_base_/base_static.py', # 定义基础配置文件路径 '../../third_party/mmdeploy/configs/_base_/backends/tensorrt-int8.py'] # 定义后端配置文件路径 onnx_config = dict(input_shape=(640, 640)) # 定义输入形状配置 backend_config = dict( # 定义后端配置 common_config=dict(max_workspace_size=1 << 30), # 定义通用配置，设置最大工作空间大小 model_inputs=[ # 定义模型输入配置 dict( # 第一个模型输入配置 input_shapes=dict( # 定义输入形状 input=dict( # 输入名称为input min_shape=[1, 3, 640, 640], # 最小形状 opt_shape=[1, 3, 640, 640], # 最佳形状 max_shape=[1, 3, 640, 640]))) # 最大形状 ]) codebase_config = dict( # 定义代码库配置 type='mmyolo', # 模型类型为mmyolo task='ObjectDetection', # 任务为目标检测 model_type='end2end', # 模型类型为端到端 post_processing=dict( # 后处理配置 score_threshold=0.1, # 分数阈值 confidence_threshold=0.005, # 置信度阈值 iou_threshold=0.3, # IoU阈值 max_output_boxes_per_class=100, # 每类最大输出框数 pre_top_k=1000, # 预选框数 keep_top_k=100, # 保留框数 background_label_id=-1), # 背景标签ID module=['mmyolo.deploy']) # 模块为mmyolo.deploy
.\YOLO-World\configs\finetune_coco\yolo_world_l_dual_vlpan_2e-4_80e_8gpus_finetune_coco.py
_base_ = ( '../../third_party/mmyolo/configs/yolov8/' 'yolov8_l_syncbn_fast_8xb16-500e_coco.py') # 定义自定义的导入模块和设置是否允许导入失败 custom_imports = dict( imports=['yolo_world'], allow_failed_imports=False) # 超参数设置 num_classes = 80 num_training_classes = 80 max_epochs = 80 # 最大训练轮数 close_mosaic_epochs = 10 save_epoch_intervals = 5 text_channels = 512 neck_embed_channels = [128, 256, _base_.last_stage_out_channels // 2] neck_num_heads = [4, 8, _base_.last_stage_out_channels // 2 // 32] base_lr = 2e-4 weight_decay = 0.05 train_batch_size_per_gpu = 16 load_from='pretrained_models/yolo_world_l_clip_base_dual_vlpan_2e-3adamw_32xb16_100e_o365_goldg_train_pretrained-0e566235.pth' persistent_workers = False # 模型设置 model = dict( type='YOLOWorldDetector', mm_neck=True, num_train_classes=num_training_classes, num_test_classes=num_classes, data_preprocessor=dict(type='YOLOWDetDataPreprocessor'), backbone=dict( _delete_=True, type='MultiModalYOLOBackbone', image_model={{_base_.model.backbone}}, text_model=dict( type='HuggingCLIPLanguageBackbone', model_name='openai/clip-vit-base-patch32', frozen_modules=['all'])), neck=dict(type='YOLOWorldDualPAFPN', guide_channels=text_channels, embed_channels=neck_embed_channels, num_heads=neck_num_heads, block_cfg=dict(type='MaxSigmoidCSPLayerWithTwoConv'), text_enhancder=dict(type='ImagePoolingAttentionModule', embed_channels=256, num_heads=8)), bbox_head=dict(type='YOLOWorldHead', head_module=dict(type='YOLOWorldHeadModule', embed_dims=text_channels, num_classes=num_training_classes)), train_cfg=dict(assigner=dict(num_classes=num_training_classes))) # 数据集设置 text_transform = [ # 定义一个字典，包含参数 type、num_neg_samples、max_num_samples、padding_to_max 和 padding_value dict(type='RandomLoadText', num_neg_samples=(num_classes, num_classes), max_num_samples=num_training_classes, padding_to_max=True, padding_value=''), # 定义一个字典，包含参数 type 和 meta_keys dict(type='mmdet.PackDetInputs', meta_keys=('img_id', 'img_path', 'ori_shape', 'img_shape', 'flip', 'flip_direction', 'texts')) # 定义一个包含多个元素的列表，每个元素是一个字典，用于进行仿射变换 mosaic_affine_transform = [ dict( type='MultiModalMosaic', img_scale=_base_.img_scale, pad_val=114.0, pre_transform=_base_.pre_transform), dict( type='YOLOv5RandomAffine', max_rotate_degree=0.0, max_shear_degree=0.0, max_aspect_ratio=100., scaling_ratio_range=(1 - _base_.affine_scale, 1 + _base_.affine_scale), # img_scale is (width, height) border=(-_base_.img_scale[0] // 2, -_base_.img_scale[1] // 2), border_val=(114, 114, 114)) ] # 定义训练数据处理流程的列表 train_pipeline = [ *_base_.pre_transform, *mosaic_affine_transform, dict( type='YOLOv5MultiModalMixUp', prob=_base_.mixup_prob, pre_transform=[*_base_.pre_transform, *mosaic_affine_transform]), *_base_.last_transform[:-1], *text_transform ] # 定义第二阶段训练数据处理流程的列表 train_pipeline_stage2 = [ *_base_.train_pipeline_stage2[:-1], *text_transform ] # 定义 COCO 训练数据集的配置字典 coco_train_dataset = dict( _delete_=True, type='MultiModalDataset', dataset=dict( type='YOLOv5CocoDataset', data_root='data/coco', ann_file='annotations/instances_train2017.json', data_prefix=dict(img='train2017/'), filter_cfg=dict(filter_empty_gt=False, min_size=32)), class_text_path='data/texts/coco_class_texts.json', pipeline=train_pipeline) # 定义训练数据加载器的配置字典 train_dataloader = dict( persistent_workers=persistent_workers, batch_size=train_batch_size_per_gpu, collate_fn=dict(type='yolow_collate'), dataset=coco_train_dataset) # 定义测试数据处理流程的列表 test_pipeline = [ *_base_.test_pipeline[:-1], dict(type='LoadText'), dict( type='mmdet.PackDetInputs', meta_keys=('img_id', 'img_path', 'ori_shape', 'img_shape', 'scale_factor', 'pad_param', 'texts')) ] # 定义 COCO 验证数据集的配置字典 coco_val_dataset = dict( _delete_=True, type='MultiModalDataset', # 定义数据集参数，指定数据集类型为YOLOv5CocoDataset dataset=dict( type='YOLOv5CocoDataset', # 数据集根目录 data_root='data/coco', # 标注文件路径 ann_file='annotations/instances_val2017.json', # 数据前缀，包含图片路径 data_prefix=dict(img='val2017/'), # 过滤配置，设置不过滤空的ground truth，最小尺寸为32 filter_cfg=dict(filter_empty_gt=False, min_size=32)), # 类别文本路径 class_text_path='data/texts/coco_class_texts.json', # 测试数据处理管道 pipeline=test_pipeline) # 创建验证数据加载器，使用 COCO 验证数据集 val_dataloader = dict(dataset=coco_val_dataset) # 将验证数据加载器赋值给测试数据加载器 test_dataloader = val_dataloader # 训练设置 # 默认钩子函数设置 default_hooks = dict( param_scheduler=dict( scheduler_type='linear', lr_factor=0.01, max_epochs=max_epochs), checkpoint=dict( max_keep_ckpts=-1, save_best=None, interval=save_epoch_intervals)) # 自定义钩子函数设置 custom_hooks = [ dict( type='EMAHook', ema_type='ExpMomentumEMA', momentum=0.0001, update_buffers=True, strict_load=False, priority=49), dict( type='mmdet.PipelineSwitchHook', switch_epoch=max_epochs - close_mosaic_epochs, switch_pipeline=train_pipeline_stage2) ] # 训练配置设置 train_cfg = dict( max_epochs=max_epochs, val_interval=5, dynamic_intervals=[((max_epochs - close_mosaic_epochs), _base_.val_interval_stage2)]) # 优化器包装器设置 optim_wrapper = dict( optimizer=dict( _delete_=True, type='AdamW', lr=base_lr, weight_decay=weight_decay, batch_size_per_gpu=train_batch_size_per_gpu), paramwise_cfg=dict( bias_decay_mult=0.0, norm_decay_mult=0.0, custom_keys={'backbone.text_model': dict(lr_mult=0.01), 'logit_scale': dict(weight_decay=0.0)}), constructor='YOLOWv5OptimizerConstructor') # 评估设置 # 验证评估器设置 val_evaluator = dict( _delete_=True, type='mmdet.CocoMetric', proposal_nums=(100, 1, 10), ann_file='data/coco/annotations/instances_val2017.json', metric='bbox')
yolo-world 源码解析（一）（2）https://developer.aliyun.com/article/1483828
