提交 2e48a7ed 编写于 作者: C cuicheng01

add PPHGNet_base config

上级 6cb33220
......@@ -133,6 +133,8 @@ PP-LCNet 系列模型的精度、速度指标如下表所示,更多关于该
**: 基于 Intel-Xeon-Gold-6271C 硬件平台与 OpenVINO 2021.4.2 推理平台。
<a name="PPHGNet"></a>
## PP-HGNet 系列
PP-HGNet 系列模型的精度、速度指标如下表所示,更多关于该系列的模型介绍可以参考:[PP-HGNet 系列模型文档](../models/PP-HGNet.md)
......@@ -140,7 +142,10 @@ PP-HGNet 系列模型的精度、速度指标如下表所示,更多关于该
| 模型 | Top-1 Acc | Top-5 Acc | time(ms)<br>bs=1 | time(ms)<br>bs=4 | time(ms)<br/>bs=8 | FLOPs(G) | Params(M) | 预训练模型下载地址 | inference模型下载地址 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| PPHGNet_tiny | 0.7983 | 0.9504 | 1.77 | - | - | 4.54 | 14.75 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/PPHGNet_tiny_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/PPHGNet_tiny_infer.tar) |
| PPHGNet_tiny_ssld | 0.8195 | 0.9612 | 1.77 | - | - | 4.54 | 14.75 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/PPHGNet_tiny_ssld_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/PPHGNet_tiny_ssld_infer.tar) |
| PPHGNet_small | 0.8151 | 0.9582 | 2.52 | - | - | 8.53 | 24.38 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/PPHGNet_small_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/PPHGNet_small_infer.tar) |
| PPHGNet_small_ssld | 0.8382 | 0.9681 | 2.52 | - | - | 8.53 | 24.38 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/PPHGNet_small_ssld_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/PPHGNet_small_ssld_infer.tar) |
| PPHGNet_base_ssld | 0.8500 | 0.9735 | 5.97 | - | - | 25.14 | 71.62 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/PPHGNet_base_ssld_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/PPHGNet_base_ssld_infer.tar) |
<a name="ResNet"></a>
......
......@@ -46,6 +46,11 @@ PP-HGNet 与其他模型的比较如下,其中测试机器为 NVIDIA® Tesla®
| SwinTransformer_tiny | 81.2 | 95.5 | 6.59 |
| <b>PPHGNet_small<b> | <b>81.51<b>| <b>95.82<b> | <b>2.52<b> |
| <b>PPHGNet_small_ssld<b> | <b>83.82<b>| <b>96.81<b> | <b>2.52<b> |
| Res2Net200_vd_26w_4s_ssld| 85.13 | 97.42 | 11.45 |
| ResNeXt101_32x48d_wsl | 85.37 | 97.69 | 55.07 |
| SwinTransformer_base | 85.2 | 97.5 | 13.53 |
| <b>PPHGNet_base_ssld<b> | <b>85.00<b>| <b>97.35<b> | <b>5.97<b> |
关于更多 PP-HGNet 的介绍以及下游任务的表现,敬请期待。
......@@ -345,7 +345,7 @@ def PPHGNet_small(pretrained=False, use_ssld=False, **kwargs):
return model
def PPHGNet_base(pretrained=False, use_ssld=False, **kwargs):
def PPHGNet_base(pretrained=False, use_ssld=True, **kwargs):
"""
PPHGNet_base
Args:
......
# global configs
Global:
checkpoints: null
pretrained_model: null
output_dir: ./output/
device: gpu
save_interval: 1
eval_during_train: True
eval_interval: 1
epochs: 600
print_batch_step: 10
use_visualdl: False
# used for static mode and model export
image_shape: [3, 224, 224]
save_inference_dir: ./inference
# training model under @to_static
to_static: False
use_dali: False
# mixed precision training
AMP:
scale_loss: 128.0
use_dynamic_loss_scaling: True
# O1: mixed fp16
level: O1
# model architecture
Arch:
name: PPHGNet_base
class_num: 1000
# loss function config for traing/eval process
Loss:
Train:
- CELoss:
weight: 1.0
epsilon: 0.1
Eval:
- CELoss:
weight: 1.0
Optimizer:
name: Momentum
momentum: 0.9
lr:
name: Cosine
learning_rate: 0.5
warmup_epoch: 5
regularizer:
name: 'L2'
coeff: 0.00004
# data loader for train and eval
DataLoader:
Train:
dataset:
name: ImageNetDataset
image_root: ./dataset/ILSVRC2012/
cls_label_path: ./dataset/ILSVRC2012/train_list.txt
transform_ops:
- DecodeImage:
to_rgb: True
channel_first: False
- RandCropImage:
size: 224
interpolation: bicubic
backend: pil
- RandFlipImage:
flip_code: 1
- TimmAutoAugment:
config_str: rand-m15-mstd0.5-inc1
interpolation: bicubic
img_size: 224
- NormalizeImage:
scale: 1.0/255.0
mean: [0.485, 0.456, 0.406]
std: [0.229, 0.224, 0.225]
order: ''
- RandomErasing:
EPSILON: 0.4
sl: 0.02
sh: 1.0/3.0
r1: 0.3
attempt: 10
use_log_aspect: True
mode: pixel
batch_transform_ops:
- OpSampler:
MixupOperator:
alpha: 0.4
prob: 0.5
CutmixOperator:
alpha: 1.0
prob: 0.5
sampler:
name: DistributedBatchSampler
batch_size: 128
drop_last: False
shuffle: True
loader:
num_workers: 16
use_shared_memory: True
Eval:
dataset:
name: ImageNetDataset
image_root: ./dataset/ILSVRC2012/
cls_label_path: ./dataset/ILSVRC2012/val_list.txt
transform_ops:
- DecodeImage:
to_rgb: True
channel_first: False
- ResizeImage:
resize_short: 236
interpolation: bicubic
backend: pil
- CropImage:
size: 224
- NormalizeImage:
scale: 1.0/255.0
mean: [0.485, 0.456, 0.406]
std: [0.229, 0.224, 0.225]
order: ''
sampler:
name: DistributedBatchSampler
batch_size: 128
drop_last: False
shuffle: False
loader:
num_workers: 16
use_shared_memory: True
Infer:
infer_imgs: docs/images/inference_deployment/whl_demo.jpg
batch_size: 10
transforms:
- DecodeImage:
to_rgb: True
channel_first: False
- ResizeImage:
resize_short: 236
- CropImage:
size: 224
- NormalizeImage:
scale: 1.0/255.0
mean: [0.485, 0.456, 0.406]
std: [0.229, 0.224, 0.225]
order: ''
- ToCHWImage:
PostProcess:
name: Topk
topk: 5
class_id_map_file: ppcls/utils/imagenet1k_label_list.txt
Metric:
Train:
- TopkAcc:
topk: [1, 5]
Eval:
- TopkAcc:
topk: [1, 5]
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册