moco-v3/vits.py

# Copyright (c) Facebook, Inc. and its affiliates.
# All rights reserved.

# This source code is licensed under the license found in the
# LICENSE file in the root directory of this source tree.

import torch
import torch.nn as nn
from functools import partial

from timm.models.vision_transformer import VisionTransformer, _cfg
from timm.models.registry import register_model

__all__ = [
    'vit_small', 
    'vit_base',
    'vit_large',
]


def vit_small(pretrained=False, **kwargs):
    model = VisionTransformer(
        patch_size=16, embed_dim=384, depth=12, num_heads=12, mlp_ratio=4, qkv_bias=True,
        norm_layer=partial(nn.LayerNorm, eps=1e-6), **kwargs)
    model.default_cfg = _cfg()
    return model


def vit_base(pretrained=False, **kwargs):
    model = VisionTransformer(
        patch_size=16, embed_dim=768, depth=12, num_heads=12, mlp_ratio=4, qkv_bias=True,
        norm_layer=partial(nn.LayerNorm, eps=1e-6), **kwargs)
    model.default_cfg = _cfg()
    return model


def vit_large(pretrained=False, **kwargs):
    model = VisionTransformer(
        patch_size=16, embed_dim=1024, depth=24, num_heads=16, mlp_ratio=4, qkv_bias=True,
        norm_layer=partial(nn.LayerNorm, eps=1e-6), **kwargs)
    model.default_cfg = _cfg()
    return model


def vit_huge(pretrained=False, **kwargs):
    model = VisionTransformer(
        patch_size=16, embed_dim=1280, depth=32, num_heads=16, mlp_ratio=4, qkv_bias=True,
        norm_layer=partial(nn.LayerNorm, eps=1e-6), **kwargs)
    model.default_cfg = _cfg()
    return model