training log

[kobe233333]

If i can get value of AP or loss on train data when i do the training, just like the log information i found in [https://download.openmmlab.com/mmpose/bottom_up/higher_hrnet32_coco_512x512_20200713.log.json] ?
Therefore i can see if i get overfit result.

[ly015]

An easy approach would be using training data as the validation set. You can set it at the data part of the config file, like:

# part of a config

data = dict(
    samples_per_gpu=32,
    workers_per_gpu=4,
    train=dict(
        type='TopDownCocoDataset',
        ann_file=f'{data_root}/annotations/person_keypoints_train2017.json',
        img_prefix=f'{data_root}/train2017/',
        data_cfg=data_cfg,
        pipeline=train_pipeline),

   # set val as your training data

    val=dict(
        type='TopDownCocoDataset',
        ann_file=f'{data_root}/annotations/person_keypoints_val2017.json',
        img_prefix=f'{data_root}/val2017/',
        data_cfg=data_cfg,
        pipeline=val_pipeline),
    test=dict(
        type='TopDownCocoDataset',
        ann_file=f'{data_root}/annotations/person_keypoints_val2017.json',
        img_prefix=f'{data_root}/val2017/',
        data_cfg=data_cfg,
        pipeline=val_pipeline),
)

[kobe233333]

An easy approach would be using training data as the validation set. You can set it at the data part of the config file, like:

# part of a config

data = dict(
    samples_per_gpu=32,
    workers_per_gpu=4,
    train=dict(
        type='TopDownCocoDataset',
        ann_file=f'{data_root}/annotations/person_keypoints_train2017.json',
        img_prefix=f'{data_root}/train2017/',
        data_cfg=data_cfg,
        pipeline=train_pipeline),

   # set val as your training data

    val=dict(
        type='TopDownCocoDataset',
        ann_file=f'{data_root}/annotations/person_keypoints_val2017.json',
        img_prefix=f'{data_root}/val2017/',
        data_cfg=data_cfg,
        pipeline=val_pipeline),
    test=dict(
        type='TopDownCocoDataset',
        ann_file=f'{data_root}/annotations/person_keypoints_val2017.json',
        img_prefix=f'{data_root}/val2017/',
        data_cfg=data_cfg,
        pipeline=val_pipeline),
)

but how i can find the loss in training process ?

[jin-s13]

The log will be saved in the work_dirs. Do you use a very small dataset?
You may have to use a smaller log interval.

mmpose/configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/res50_coco_256x192.py

Line 23 in 7d9b249

interval=50,

[kobe233333]

The log will be saved in the work_dirs. Do you use a very small dataset?
You may have to use a smaller log interval.

mmpose/configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/res50_coco_256x192.py

Line 23 in 7d9b249

interval=50,

but i can only see the ap score, no training loss

[jin-s13]

Please try interval = 1

[jin-s13]

Have you solved this problem?

[kobe233333]

Have you solved this problem?

yes, appreciate!