mmclassification/configs/eva/eva-g-p14_headless.py
takuoko 2535c1ecd7
[Feature] Support EVA. (#1239)
* add eva

* add eva

* add eva

* sklearn -> scikit-learn

* add large

* Update model names and links.

* Fix resize pos embed error when loading fp16 weight.

* Remove verbose configs.

Co-authored-by: mzr1996 <mzr1996@163.com>
2022-12-14 13:21:33 +08:00

26 lines
633 B
Python

model = dict(
type='ImageClassifier',
backbone=dict(
type='BEiT',
arch='eva-g',
img_size=224,
patch_size=14,
avg_token=True,
layer_scale_init_value=0.0,
output_cls_token=False,
use_abs_pos_emb=True,
use_rel_pos_bias=False,
use_shared_rel_pos_bias=False,
),
neck=None,
head=None,
)
data_preprocessor = dict(
# RGB format normalization parameters
mean=[0.48145466 * 255, 0.4578275 * 255, 0.40821073 * 255],
std=[0.26862954 * 255, 0.26130258 * 255, 0.27577711 * 255],
# convert image from BGR to RGB
to_rgb=True,
)