mirrors
/
PaddleClas
mirror of https://github.com/PaddlePaddle/PaddleClas.git
1
0
Fork 0
Code Issues Projects Releases Wiki Activity

update foundation_vit from EVA_vit_huge to EVA_vit_giant

pull/2765/head
zh-hike 2023-04-14 02:30:57 +00:00 committed by zengshao0622
parent 0af4680f86
commit d7bd275379
3 changed files with 23 additions and 23 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

38
docs/zh_CN/models/Foundation_models/FoundationViT.md
View File

@ -13,7 +13,7 @@
## 2. 使用说明
以模型`CLIP_vit_base_patch16_224`为例，使用该模型以及对应的预训练权重进行特征提取的代码如下：
以模型 `CLIP_vit_base_patch16_224`为例，使用该模型以及对应的预训练权重进行特征提取的代码如下：
```python
from ppcls.utils import config
@ -31,23 +31,23 @@ output = model(inputs) # the output of model embeding
目前支持的视觉大模型以及预训练权重如下：
| 系列 | 模型 | 模型大小 | embedding_size | 预训练数据集 | 权重下载 |
| :----: | :--------------------------------: | :------: | :------------: | :----------------------------------------------: | ------------------------------------------------------------ |
| CLIP | CLIP_vit_base_patch16_224 | 85M | 768 | WIT | [下载地址](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/foundation_models/CLIP_vit_base_patch16_224.pdparams) |
| CLIP | CLIP_vit_base_patch32_224 | 87M | 768 | WIT | [下载地址](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/foundation_models/CLIP_vit_base_patch32_224.pdparams) |
| CLIP | CLIP_vit_large_patch14_224 | 302M | 1024 | WIT | [下载地址](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/foundation_models/CLIP_vit_large_patch14_224.pdparams) |
| CLIP | CLIP_vit_large_patch14_336 | 302M | 1024 | WIT | [下载地址](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/foundation_models/CLIP_vit_large_patch14_336.pdparams) |
| BEiTv2 | BEiTv2_vit_base_patch16_224 | 85M | 768 | ImageNet-1k | [下载地址](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/foundation_models/BEiTv2_vit_base_patch16_224.pdparams) |
| BEiTv2 | BEiTv2_vit_base_patch16_224_ft21k | 85M | 768 | ImageNet-1k、ImageNet-21k | [下载地址](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/foundation_models/BEiTv2_vit_base_patch16_224_ft21k.pdparams) |
| BEiTv2 | BEiTv2_vit_large_patch16_224 | 303M | 1024 | ImageNet-1k | [下载地址](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/foundation_models/BEiTv2_vit_large_patch16_224.pdparams) |
| BEiTv2 | BEiTv2_vit_large_patch16_224_ft21k | 303M | 1024 | ImageNet-1k、ImageNet-21k | [下载地址](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/foundation_models/BEiTv2_vit_large_patch16_224_ft21k.pdparams) |
| MoCoV3 | MoCoV3_vit_small | 21M | 384 | ImageNet-1k | [下载地址](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/foundation_models/MoCoV3_vit_small.pdparams) |
| MoCoV3 | MoCoV3_vit_base | 85M | 768 | ImageNet-1k | [下载地址](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/foundation_models/MoCoV3_vit_base.pdparams) |
| MAE | MAE_vit_base_patch16 | 85M | 768 | ImageNet-1k | [下载地址](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/foundation_models/MAE_vit_base_patch16.pdparams) |
| MAE | MAE_vit_large_patch16 | 303M | 1024 | ImageNet-1k | [下载地址](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/foundation_models/MAE_vit_large_patch16.pdparams) |
| MAE | MAE_vit_huge_patch14 | 630M | 1280 | ImageNet-1k | [下载地址](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/foundation_models/MAE_vit_huge_patch14.pdparams) |
| EVA | EVA_vit_huge_patch14 | 1010M | 1408 | ImageNet-21k, CC12M, CC2M, Object365,COCO, ADE | [下载地址](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/foundation_models/EVA_vit_huge_patch14.pdparams) |
| CAE | CAE_vit_base_patch16_224 | 85M | 768 | ImageNet-1k | [下载地址](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/foundation_models/CAE_vit_base_patch16_224.pdparams) |
| 系列 | 模型 | 模型大小 | embedding_size | 预训练数据集 | 权重下载 |
| :----: | :--------------------------------: | :------: | :------------: | :----------------------------------------------: | -------------------------------------------------------------------------------------------------------------------------------- |
| CLIP | CLIP_vit_base_patch16_224 | 85M | 768 | WIT | [下载地址](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/foundation_models/CLIP_vit_base_patch16_224.pdparams) |
| CLIP | CLIP_vit_base_patch32_224 | 87M | 768 | WIT | [下载地址](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/foundation_models/CLIP_vit_base_patch32_224.pdparams) |
| CLIP | CLIP_vit_large_patch14_224 | 302M | 1024 | WIT | [下载地址](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/foundation_models/CLIP_vit_large_patch14_224.pdparams) |
| CLIP | CLIP_vit_large_patch14_336 | 302M | 1024 | WIT | [下载地址](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/foundation_models/CLIP_vit_large_patch14_336.pdparams) |
| BEiTv2 | BEiTv2_vit_base_patch16_224 | 85M | 768 | ImageNet-1k | [下载地址](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/foundation_models/BEiTv2_vit_base_patch16_224.pdparams) |
| BEiTv2 | BEiTv2_vit_base_patch16_224_ft21k | 85M | 768 | ImageNet-1k、ImageNet-21k | [下载地址](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/foundation_models/BEiTv2_vit_base_patch16_224_ft21k.pdparams) |
| BEiTv2 | BEiTv2_vit_large_patch16_224 | 303M | 1024 | ImageNet-1k | [下载地址](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/foundation_models/BEiTv2_vit_large_patch16_224.pdparams) |
| BEiTv2 | BEiTv2_vit_large_patch16_224_ft21k | 303M | 1024 | ImageNet-1k、ImageNet-21k | [下载地址](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/foundation_models/BEiTv2_vit_large_patch16_224_ft21k.pdparams) |
| MoCoV3 | MoCoV3_vit_small | 21M | 384 | ImageNet-1k | [下载地址](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/foundation_models/MoCoV3_vit_small.pdparams) |
| MoCoV3 | MoCoV3_vit_base | 85M | 768 | ImageNet-1k | [下载地址](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/foundation_models/MoCoV3_vit_base.pdparams) |
| MAE | MAE_vit_base_patch16 | 85M | 768 | ImageNet-1k | [下载地址](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/foundation_models/MAE_vit_base_patch16.pdparams) |
| MAE | MAE_vit_large_patch16 | 303M | 1024 | ImageNet-1k | [下载地址](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/foundation_models/MAE_vit_large_patch16.pdparams) |
| MAE | MAE_vit_huge_patch14 | 630M | 1280 | ImageNet-1k | [下载地址](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/foundation_models/MAE_vit_huge_patch14.pdparams) |
| EVA | EVA_vit_giant_patch14 | 1010M | 1408 | ImageNet-21k, CC12M, CC2M, Object365,COCO, ADE | [下载地址](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/foundation_models/EVA_vit_giant_patch14.pdparams) |
| CAE | CAE_vit_base_patch16_224 | 85M | 768 | ImageNet-1k | [下载地址](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/foundation_models/CAE_vit_base_patch16_224.pdparams) |
## 4. 参考文献
@ -56,4 +56,4 @@ output = model(inputs) # the output of model embeding
3. [BEiT v2: Masked Image Modeling with Vector-Quantized Visual Tokenizers](https://arxiv.org/abs/2208.06366)
4. [CAE: Context Autoencoder for Self-Supervised Representation Learning](https://arxiv.org/abs/2202.03026)
5. [EVA: EVA: Exploring the Limits of Masked Visual Representation Learning at Scale](https://paperswithcode.com/paper/eva-exploring-the-limits-of-masked-visual)
6. [MAE: Masked Autoencoders Are Scalable Vision Learners](https://paperswithcode.com/paper/masked-autoencoders-are-scalable-vision)
6. [MAE: Masked Autoencoders Are Scalable Vision Learners](https://paperswithcode.com/paper/masked-autoencoders-are-scalable-vision)

2
ppcls/arch/backbone/__init__.py
View File

@ -71,7 +71,7 @@ from .model_zoo.mobilevit import MobileViT_XXS, MobileViT_XS, MobileViT_S
from .model_zoo.repvgg import RepVGG_A0, RepVGG_A1, RepVGG_A2, RepVGG_B0, RepVGG_B1, RepVGG_B2, RepVGG_B1g2, RepVGG_B1g4, RepVGG_B2g4, RepVGG_B3, RepVGG_B3g4, RepVGG_D2se
from .model_zoo.van import VAN_B0, VAN_B1, VAN_B2, VAN_B3
from .model_zoo.peleenet import PeleeNet
from .model_zoo.foundation_vit import CLIP_vit_base_patch32_224, CLIP_vit_base_patch16_224, CLIP_vit_large_patch14_336, CLIP_vit_large_patch14_224, BEiTv2_vit_base_patch16_224, BEiTv2_vit_large_patch16_224, CAE_vit_base_patch16_224, EVA_vit_huge_patch14, MOCOV3_vit_small, MOCOV3_vit_base, MAE_vit_huge_patch14, MAE_vit_large_patch16, MAE_vit_base_patch16
from .model_zoo.foundation_vit import CLIP_vit_base_patch32_224, CLIP_vit_base_patch16_224, CLIP_vit_large_patch14_336, CLIP_vit_large_patch14_224, BEiTv2_vit_base_patch16_224, BEiTv2_vit_large_patch16_224, CAE_vit_base_patch16_224, EVA_vit_giant_patch14, MOCOV3_vit_small, MOCOV3_vit_base, MAE_vit_huge_patch14, MAE_vit_large_patch16, MAE_vit_base_patch16
from .model_zoo.convnext import ConvNeXt_tiny, ConvNeXt_small, ConvNeXt_base_224, ConvNeXt_base_384, ConvNeXt_large_224, ConvNeXt_large_384
from .model_zoo.nextvit import NextViT_small_224, NextViT_base_224, NextViT_large_224, NextViT_small_384, NextViT_base_384, NextViT_large_384
from .model_zoo.cae import cae_base_patch16_224, cae_large_patch16_224

6
ppcls/arch/backbone/model_zoo/foundation_vit.py
View File

@ -40,8 +40,8 @@ MODEL_URLS = {
"https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/foundation_models/BEiTv2_vit_large_patch16_224.pdparams",
"CAE_vit_base_patch16_224":
"https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/foundation_models/CAE_vit_base_patch16_224.pdparams",
'EVA_vit_huge_patch14':
"https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/foundation_models/EVA_vit_huge_patch14.pdparams",
'EVA_vit_giant_patch14':
"https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/foundation_models/EVA_vit_giant_patch14.pdparams",
"MOCOV3_vit_small":
"https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/foundation_models/MOCOV3_vit_small.pdparams",
"MOCOV3_vit_base":
@ -867,7 +867,7 @@ def MAE_vit_huge_patch14(pretrained=False, use_ssld=False, **kwargs):
return model
def EVA_vit_huge_patch14(pretrained=False, use_ssld=False, **kwargs):
def EVA_vit_giant_patch14(pretrained=False, use_ssld=False, **kwargs):
model_name = sys._getframe().f_code.co_name
model = VisionTransformer(
model_name=model_name,