PaddlePaddle
用PaddleClas训练2类模型，训练的top1精度错误。在训练过程中推理验证集和训练完成后单独推理验证集输出结果不一致。
在训练2分类模型时，训练过程中训练集和验证集上的精度都能够达到0.99， 训练完成后对验证集上的数据进行推理，精度只有0.57，请问问题是什么？
提出issue时，辛苦您提供以下信息，方便我们快速定位问题并及时有效地解决您的问题：
- PaddleClas release/2.5和PaddlePaddle 2.4.2
- 训练环境信息： a. 具体操作系统，如Windows b. Python版本号，如Python3.9 c. CUDA/cuDNN版本， 如CUDA11.2/cuDNN 7.6.5等
- 完整配置文件
global configs
Global: checkpoints: null pretrained_model: null output_dir: ./output/ device: gpu save_interval: 1 eval_during_train: True eval_interval: 1 epochs: 40 print_batch_step: 10
use_visualdl: False
used for static mode and model export
image_shape: [3, 448, 448] save_inference_dir: ./inference
model architecture
Arch: name: ShuffleNetV2_x0_25 class_num: 2
loss function config for traing/eval process
Loss: Train: - CELoss: weight: 1.0 Eval: - CELoss: weight: 1.0
Optimizer: name: Momentum momentum: 0.9 lr: name: Cosine learning_rate: 0.0025 warmup_epoch: 5 regularizer: name: 'L2' coeff: 0.00001
data loader for train and eval
DataLoader: Train: dataset: name: ImageNetDataset image_root: ./dataset/model_35/ cls_label_path: ./dataset/model_35/train.txt transform_ops: - DecodeImage: to_rgb: True channel_first: False - RandCropImage: size: 448 - RandFlipImage: flip_code: 1 - 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: 64
drop_last: False
shuffle: True
loader:
num_workers: 1
use_shared_memory: True
Eval: dataset: name: ImageNetDataset image_root: ./dataset/model_35/ cls_label_path: ./dataset/model_35/val.txt transform_ops: - DecodeImage: to_rgb: True channel_first: False - ResizeImage: resize_short: 512 - CropImage: size: 448 - 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: 16 drop_last: False shuffle: False loader: num_workers: 1 use_shared_memory: True
Infer: infer_imgs: D:\lixiaolin\PaddleClas-release-2.5\dataset\model_35\images batch_size: 8 transforms: - DecodeImage: to_rgb: True channel_first: False - ResizeImage: resize_short: 512 - CropImage: size: 448 - 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: 1 class_id_map_file: ./dataset/model_35/label_list.txt
Metric: Train: - TopkAcc: topk: [1, 1] Eval: - TopkAcc: topk: [1, 1]
补充验证集上测试结果，以下为推理过程中的， [2024/07/30 13:14:25] ppcls INFO: [Eval][Epoch 1][Iter: 0/195]CELoss: 0.52470, loss: 0.52470, top1: 0.87500, batch_cost: 0.29910s, reader_cost: 0.26422, ips: 53.49302 images/sec [2024/07/30 13:14:28] ppcls INFO: [Eval][Epoch 1][Iter: 10/195]CELoss: 0.41228, loss: 0.41228, top1: 0.93750, batch_cost: 0.25465s, reader_cost: 0.22881, ips: 62.83192 images/sec [2024/07/30 13:14:31] ppcls INFO: [Eval][Epoch 1][Iter: 20/195]CELoss: 0.39574, loss: 0.39574, top1: 0.94643, batch_cost: 0.27717s, reader_cost: 0.22224, ips: 57.72577 images/sec [2024/07/30 13:14:34] ppcls INFO: [Eval][Epoch 1][Iter: 30/195]CELoss: 0.41059, loss: 0.41059, top1: 0.95565, batch_cost: 0.29538s, reader_cost: 0.21609, ips: 54.16705 images/sec [2024/07/30 13:14:37] ppcls INFO: [Eval][Epoch 1][Iter: 40/195]CELoss: 0.49828, loss: 0.49828, top1: 0.95122, batch_cost: 0.30362s, reader_cost: 0.21354, ips: 52.69826 images/sec [2024/07/30 13:14:40] ppcls INFO: [Eval][Epoch 1][Iter: 50/195]CELoss: 0.53395, loss: 0.53395, top1: 0.94485, batch_cost: 0.30908s, reader_cost: 0.21277, ips: 51.76588 images/sec [2024/07/30 13:14:44] ppcls INFO: [Eval][Epoch 1][Iter: 60/195]CELoss: 0.39273, loss: 0.39273, top1: 0.94877, batch_cost: 0.31239s, reader_cost: 0.21210, ips: 51.21774 images/sec [2024/07/30 13:14:47] ppcls INFO: [Eval][Epoch 1][Iter: 70/195]CELoss: 0.57583, loss: 0.57583, top1: 0.90757, batch_cost: 0.31467s, reader_cost: 0.21168, ips: 50.84639 images/sec [2024/07/30 13:14:50] ppcls INFO: [Eval][Epoch 1][Iter: 80/195]CELoss: 0.76076, loss: 0.76076, top1: 0.87037, batch_cost: 0.31643s, reader_cost: 0.21150, ips: 50.56428 images/sec [2024/07/30 13:14:53] ppcls INFO: [Eval][Epoch 1][Iter: 90/195]CELoss: 0.65214, loss: 0.65214, top1: 0.81181, batch_cost: 0.31836s, reader_cost: 0.21180, ips: 50.25713 images/sec [2024/07/30 13:14:57] ppcls INFO: [Eval][Epoch 1][Iter: 100/195]CELoss: 0.74966, loss: 0.74966, top1: 0.77785, batch_cost: 0.31696s, reader_cost: 0.20912, ips: 50.47984 images/sec [2024/07/30 13:14:59] ppcls INFO: [Eval][Epoch 1][Iter: 110/195]CELoss: 0.71694, loss: 0.71694, top1: 0.74718, batch_cost: 0.31484s, reader_cost: 0.20591, ips: 50.81927 images/sec [2024/07/30 13:15:02] ppcls INFO: [Eval][Epoch 1][Iter: 120/195]CELoss: 0.77488, loss: 0.77488, top1: 0.71488, batch_cost: 0.31315s, reader_cost: 0.20337, ips: 51.09332 images/sec [2024/07/30 13:15:05] ppcls INFO: [Eval][Epoch 1][Iter: 130/195]CELoss: 0.74157, loss: 0.74157, top1: 0.68702, batch_cost: 0.31215s, reader_cost: 0.20177, ips: 51.25716 images/sec [2024/07/30 13:15:08] ppcls INFO: [Eval][Epoch 1][Iter: 140/195]CELoss: 0.75262, loss: 0.75262, top1: 0.66356, batch_cost: 0.31097s, reader_cost: 0.20003, ips: 51.45263 images/sec [2024/07/30 13:15:11] ppcls INFO: [Eval][Epoch 1][Iter: 150/195]CELoss: 0.77913, loss: 0.77913, top1: 0.64735, batch_cost: 0.30926s, reader_cost: 0.19778, ips: 51.73663 images/sec [2024/07/30 13:15:14] ppcls INFO: [Eval][Epoch 1][Iter: 160/195]CELoss: 0.65055, loss: 0.65055, top1: 0.63121, batch_cost: 0.30774s, reader_cost: 0.19584, ips: 51.99117 images/sec [2024/07/30 13:15:17] ppcls INFO: [Eval][Epoch 1][Iter: 170/195]CELoss: 0.70114, loss: 0.70114, top1: 0.61513, batch_cost: 0.30651s, reader_cost: 0.19419, ips: 52.20137 images/sec [2024/07/30 13:15:20] ppcls INFO: [Eval][Epoch 1][Iter: 180/195]CELoss: 0.78585, loss: 0.78585, top1: 0.60221, batch_cost: 0.30573s, reader_cost: 0.19305, ips: 52.33291 images/sec [2024/07/30 13:15:23] ppcls INFO: [Eval][Epoch 1][Iter: 190/195]CELoss: 0.68863, loss: 0.68863, top1: 0.59130, batch_cost: 0.30481s, reader_cost: 0.19180, ips: 52.49220 images/sec [2024/07/30 13:15:24] ppcls INFO: [Eval][Epoch 1][Avg]CELoss: 0.62258, loss: 0.62258, top1: 0.58763 以下为单独推理的， W0730 13:22:11.998972 16512 gpu_resources.cc:91] device: 0, cuDNN Version: 8.2. [2024/07/30 13:22:15] ppcls INFO: [Eval][Epoch 0][Iter: 0/195]CELoss: 0.00000, loss: 0.00000, top1: 1.00000, batch_cost: 2.50228s, reader_cost: 0.25417, ips: 6.39418 images/sec [2024/07/30 13:22:16] ppcls INFO: [Eval][Epoch 0][Iter: 10/195]CELoss: 0.00000, loss: 0.00000, top1: 1.00000, batch_cost: 0.20145s, reader_cost: 0.18583, ips: 79.42415 images/sec [2024/07/30 13:22:19] ppcls INFO: [Eval][Epoch 0][Iter: 20/195]CELoss: 0.00000, loss: 0.00000, top1: 1.00000, batch_cost: 0.23395s, reader_cost: 0.21145, ips: 68.39063 images/sec [2024/07/30 13:22:22] ppcls INFO: [Eval][Epoch 0][Iter: 30/195]CELoss: 0.00000, loss: 0.00000, top1: 1.00000, batch_cost: 0.25897s, reader_cost: 0.21222, ips: 61.78229 images/sec [2024/07/30 13:22:25] ppcls INFO: [Eval][Epoch 0][Iter: 40/195]CELoss: 0.00000, loss: 0.00000, top1: 1.00000, batch_cost: 0.27579s, reader_cost: 0.21002, ips: 58.01453 images/sec [2024/07/30 13:22:28] ppcls INFO: [Eval][Epoch 0][Iter: 50/195]CELoss: 0.00000, loss: 0.00000, top1: 1.00000, batch_cost: 0.28785s, reader_cost: 0.21129, ips: 55.58460 images/sec [2024/07/30 13:22:31] ppcls INFO: [Eval][Epoch 0][Iter: 60/195]CELoss: 0.00000, loss: 0.00000, top1: 1.00000, batch_cost: 0.29454s, reader_cost: 0.21093, ips: 54.32187 images/sec [2024/07/30 13:22:35] ppcls INFO: [Eval][Epoch 0][Iter: 70/195]CELoss: 0.00000, loss: 0.00000, top1: 1.00000, batch_cost: 0.29824s, reader_cost: 0.20981, ips: 53.64878 images/sec [2024/07/30 13:22:38] ppcls INFO: [Eval][Epoch 0][Iter: 80/195]CELoss: 0.00000, loss: 0.00000, top1: 1.00000, batch_cost: 0.30161s, reader_cost: 0.20971, ips: 53.04938 images/sec [2024/07/30 13:22:41] ppcls INFO: [Eval][Epoch 0][Iter: 90/195]CELoss: 0.00000, loss: 0.00000, top1: 1.00000, batch_cost: 0.30515s, reader_cost: 0.21053, ips: 52.43406 images/sec [2024/07/30 13:22:44] ppcls INFO: [Eval][Epoch 0][Iter: 100/195]CELoss: 129.15761, loss: 129.15761, top1: 0.91275, batch_cost: 0.30472s, reader_cost: 0.20793, ips: 52.50744 images/sec [2024/07/30 13:22:47] ppcls INFO: [Eval][Epoch 0][Iter: 110/195]CELoss: 134.15138, loss: 134.15138, top1: 0.83052, batch_cost: 0.30310s, reader_cost: 0.20454, ips: 52.78798 images/sec [2024/07/30 13:22:50] ppcls INFO: [Eval][Epoch 0][Iter: 120/195]CELoss: 140.31223, loss: 140.31223, top1: 0.76188, batch_cost: 0.30127s, reader_cost: 0.20128, ips: 53.10837 images/sec [2024/07/30 13:22:53] ppcls INFO: [Eval][Epoch 0][Iter: 130/195]CELoss: 128.21762, loss: 128.21762, top1: 0.70372, batch_cost: 0.30011s, reader_cost: 0.19891, ips: 53.31406 images/sec [2024/07/30 13:22:56] ppcls INFO: [Eval][Epoch 0][Iter: 140/195]CELoss: 129.90176, loss: 129.90176, top1: 0.65381, batch_cost: 0.29914s, reader_cost: 0.19691, ips: 53.48637 images/sec [2024/07/30 13:22:59] ppcls INFO: [Eval][Epoch 0][Iter: 150/195]CELoss: 134.15936, loss: 134.15936, top1: 0.61051, batch_cost: 0.29811s, reader_cost: 0.19501, ips: 53.67186 images/sec [2024/07/30 13:23:01] ppcls INFO: [Eval][Epoch 0][Iter: 160/195]CELoss: 144.06953, loss: 144.06953, top1: 0.57259, batch_cost: 0.29712s, reader_cost: 0.19324, ips: 53.85047 images/sec [2024/07/30 13:23:04] ppcls INFO: [Eval][Epoch 0][Iter: 170/195]CELoss: 142.10345, loss: 142.10345, top1: 0.53911, batch_cost: 0.29655s, reader_cost: 0.19196, ips: 53.95314 images/sec [2024/07/30 13:23:07] ppcls INFO: [Eval][Epoch 0][Iter: 180/195]CELoss: 128.49042, loss: 128.49042, top1: 0.50932, batch_cost: 0.29585s, reader_cost: 0.19064, ips: 54.08175 images/sec [2024/07/30 13:23:10] ppcls INFO: [Eval][Epoch 0][Iter: 190/195]CELoss: 125.30135, loss: 125.30135, top1: 0.48266, batch_cost: 0.29518s, reader_cost: 0.18941, ips: 54.20476 images/sec [2024/07/30 13:23:11] ppcls INFO: [Eval][Epoch 0][Avg]CELoss: 71.23187, loss: 71.23187, top1: 0.47519
只训练了1个epoch的，用了resnet50，麻烦再给看下原因 python tools/train.py -c D:\lixiaolin\PaddleClas-release-2.5\ppcls\configs\ImageNet\ResNet\ResNet50.yaml train.log python tools/eval.py -c D:\lixiaolin\PaddleClas-release-2.5\ppcls\configs\ImageNet\ResNet\ResNet50.yaml -o ARCHITECTURE.name="ResNet50" -o pretrained_model=D:\lixiaolin\PaddleClas-release-2.5\output\ResNet50\best.pdparams eval.log
result.txt 我还对训练数据集进行了推理，统计结果中，错误的262个，正确的5862个，对应的分类准确率95.72%，跟eval的结果差别很大。 总结一下：训练集给的结果是0.78557，验证集在训练过程中的评估结果是0.96635，直接进行评估结果是0.48237，直接推训练集和验证集的结果是0.9572.训练集和验证集数据8:2分割。请帮忙解释一下数据的原因
还有一个问题，这个平均值是怎么算出来了，感觉怎么都不到0.98211 [2024/07/31 20:24:32] ppcls INFO: [Train][Epoch 24/120][Iter: 0/304]lr(PiecewiseDecay): 0.00100000, top1: 1.00000, CELoss: 0.01757, loss: 0.01757, batch_cost: 0.61280s, reader_cost: 0.12493, ips: 26.10981 samples/s, eta: 5:01:10 [2024/07/31 20:24:38] ppcls INFO: [Train][Epoch 24/120][Iter: 10/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97159, CELoss: 0.05098, loss: 0.05098, batch_cost: 0.62808s, reader_cost: 0.12837, ips: 25.47451 samples/s, eta: 5:08:34 [2024/07/31 20:24:44] ppcls INFO: [Train][Epoch 24/120][Iter: 20/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97619, CELoss: 0.05253, loss: 0.05253, batch_cost: 0.62088s, reader_cost: 0.12821, ips: 25.76972 samples/s, eta: 5:04:56 [2024/07/31 20:24:51] ppcls INFO: [Train][Epoch 24/120][Iter: 30/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97379, CELoss: 0.06434, loss: 0.06434, batch_cost: 0.62541s, reader_cost: 0.13184, ips: 25.58316 samples/s, eta: 5:07:03 [2024/07/31 20:24:57] ppcls INFO: [Train][Epoch 24/120][Iter: 40/304]lr(PiecewiseDecay): 0.00100000, top1: 0.96951, CELoss: 0.06668, loss: 0.06668, batch_cost: 0.62095s, reader_cost: 0.12972, ips: 25.76705 samples/s, eta: 5:04:45 [2024/07/31 20:25:03] ppcls INFO: [Train][Epoch 24/120][Iter: 50/304]lr(PiecewiseDecay): 0.00100000, top1: 0.96814, CELoss: 0.06468, loss: 0.06468, batch_cost: 0.61844s, reader_cost: 0.12773, ips: 25.87151 samples/s, eta: 5:03:25 [2024/07/31 20:25:09] ppcls INFO: [Train][Epoch 24/120][Iter: 60/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97029, CELoss: 0.06370, loss: 0.06370, batch_cost: 0.61698s, reader_cost: 0.12703, ips: 25.93262 samples/s, eta: 5:02:36 [2024/07/31 20:25:15] ppcls INFO: [Train][Epoch 24/120][Iter: 70/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97359, CELoss: 0.05873, loss: 0.05873, batch_cost: 0.61571s, reader_cost: 0.12661, ips: 25.98613 samples/s, eta: 5:01:53 [2024/07/31 20:25:21] ppcls INFO: [Train][Epoch 24/120][Iter: 80/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97531, CELoss: 0.05557, loss: 0.05557, batch_cost: 0.61537s, reader_cost: 0.12652, ips: 26.00050 samples/s, eta: 5:01:36 [2024/07/31 20:25:27] ppcls INFO: [Train][Epoch 24/120][Iter: 90/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97734, CELoss: 0.05107, loss: 0.05107, batch_cost: 0.61270s, reader_cost: 0.12442, ips: 26.11384 samples/s, eta: 5:00:12 [2024/07/31 20:25:33] ppcls INFO: [Train][Epoch 24/120][Iter: 100/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97772, CELoss: 0.05041, loss: 0.05041, batch_cost: 0.61239s, reader_cost: 0.12465, ips: 26.12733 samples/s, eta: 4:59:56 [2024/07/31 20:25:39] ppcls INFO: [Train][Epoch 24/120][Iter: 110/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97804, CELoss: 0.05052, loss: 0.05052, batch_cost: 0.61104s, reader_cost: 0.12344, ips: 26.18475 samples/s, eta: 4:59:11 [2024/07/31 20:25:45] ppcls INFO: [Train][Epoch 24/120][Iter: 120/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97882, CELoss: 0.04917, loss: 0.04917, batch_cost: 0.60958s, reader_cost: 0.12228, ips: 26.24745 samples/s, eta: 4:58:22 [2024/07/31 20:25:51] ppcls INFO: [Train][Epoch 24/120][Iter: 130/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97996, CELoss: 0.04716, loss: 0.04716, batch_cost: 0.60957s, reader_cost: 0.12241, ips: 26.24819 samples/s, eta: 4:58:15 [2024/07/31 20:25:57] ppcls INFO: [Train][Epoch 24/120][Iter: 140/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97961, CELoss: 0.05031, loss: 0.05031, batch_cost: 0.60854s, reader_cost: 0.12157, ips: 26.29241 samples/s, eta: 4:57:39 [2024/07/31 20:26:03] ppcls INFO: [Train][Epoch 24/120][Iter: 150/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97972, CELoss: 0.04904, loss: 0.04904, batch_cost: 0.60852s, reader_cost: 0.12143, ips: 26.29311 samples/s, eta: 4:57:32 [2024/07/31 20:26:10] ppcls INFO: [Train][Epoch 24/120][Iter: 160/304]lr(PiecewiseDecay): 0.00100000, top1: 0.98059, CELoss: 0.04801, loss: 0.04801, batch_cost: 0.60879s, reader_cost: 0.12156, ips: 26.28154 samples/s, eta: 4:57:34 [2024/07/31 20:26:15] ppcls INFO: [Train][Epoch 24/120][Iter: 170/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97990, CELoss: 0.04968, loss: 0.04968, batch_cost: 0.60820s, reader_cost: 0.12103, ips: 26.30717 samples/s, eta: 4:57:11 [2024/07/31 20:26:22] ppcls INFO: [Train][Epoch 24/120][Iter: 180/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97963, CELoss: 0.04977, loss: 0.04977, batch_cost: 0.60820s, reader_cost: 0.12117, ips: 26.30719 samples/s, eta: 4:57:05 [2024/07/31 20:26:28] ppcls INFO: [Train][Epoch 24/120][Iter: 190/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97971, CELoss: 0.04987, loss: 0.04987, batch_cost: 0.60803s, reader_cost: 0.12091, ips: 26.31468 samples/s, eta: 4:56:53 [2024/07/31 20:26:34] ppcls INFO: [Train][Epoch 24/120][Iter: 200/304]lr(PiecewiseDecay): 0.00100000, top1: 0.98010, CELoss: 0.05034, loss: 0.05034, batch_cost: 0.60766s, reader_cost: 0.12063, ips: 26.33051 samples/s, eta: 4:56:37 [2024/07/31 20:26:40] ppcls INFO: [Train][Epoch 24/120][Iter: 210/304]lr(PiecewiseDecay): 0.00100000, top1: 0.98015, CELoss: 0.05039, loss: 0.05039, batch_cost: 0.60726s, reader_cost: 0.12033, ips: 26.34805 samples/s, eta: 4:56:19 [2024/07/31 20:26:46] ppcls INFO: [Train][Epoch 24/120][Iter: 220/304]lr(PiecewiseDecay): 0.00100000, top1: 0.98077, CELoss: 0.04899, loss: 0.04899, batch_cost: 0.60762s, reader_cost: 0.12083, ips: 26.33223 samples/s, eta: 4:56:23 [2024/07/31 20:26:52] ppcls INFO: [Train][Epoch 24/120][Iter: 230/304]lr(PiecewiseDecay): 0.00100000, top1: 0.98079, CELoss: 0.04874, loss: 0.04874, batch_cost: 0.60726s, reader_cost: 0.12061, ips: 26.34782 samples/s, eta: 4:56:07 [2024/07/31 20:26:58] ppcls INFO: [Train][Epoch 24/120][Iter: 240/304]lr(PiecewiseDecay): 0.00100000, top1: 0.98107, CELoss: 0.04837, loss: 0.04837, batch_cost: 0.60674s, reader_cost: 0.12025, ips: 26.37032 samples/s, eta: 4:55:46 [2024/07/31 20:27:04] ppcls INFO: [Train][Epoch 24/120][Iter: 250/304]lr(PiecewiseDecay): 0.00100000, top1: 0.98157, CELoss: 0.04781, loss: 0.04781, batch_cost: 0.60659s, reader_cost: 0.12022, ips: 26.37714 samples/s, eta: 4:55:35 [2024/07/31 20:27:10] ppcls INFO: [Train][Epoch 24/120][Iter: 260/304]lr(PiecewiseDecay): 0.00100000, top1: 0.98156, CELoss: 0.04710, loss: 0.04710, batch_cost: 0.60626s, reader_cost: 0.12001, ips: 26.39147 samples/s, eta: 4:55:19 [2024/07/31 20:27:16] ppcls INFO: [Train][Epoch 24/120][Iter: 270/304]lr(PiecewiseDecay): 0.00100000, top1: 0.98155, CELoss: 0.04814, loss: 0.04814, batch_cost: 0.60576s, reader_cost: 0.11958, ips: 26.41314 samples/s, eta: 4:54:59 [2024/07/31 20:27:22] ppcls INFO: [Train][Epoch 24/120][Iter: 280/304]lr(PiecewiseDecay): 0.00100000, top1: 0.98132, CELoss: 0.04887, loss: 0.04887, batch_cost: 0.60638s, reader_cost: 0.12027, ips: 26.38607 samples/s, eta: 4:55:11 [2024/07/31 20:27:28] ppcls INFO: [Train][Epoch 24/120][Iter: 290/304]lr(PiecewiseDecay): 0.00100000, top1: 0.98174, CELoss: 0.04826, loss: 0.04826, batch_cost: 0.60602s, reader_cost: 0.11995, ips: 26.40185 samples/s, eta: 4:54:54 [2024/07/31 20:27:34] ppcls INFO: [Train][Epoch 24/120][Iter: 300/304]lr(PiecewiseDecay): 0.00100000, top1: 0.98214, CELoss: 0.04717, loss: 0.04717, batch_cost: 0.60588s, reader_cost: 0.11982, ips: 26.40767 samples/s, eta: 4:54:44 [2024/07/31 20:27:36] ppcls INFO: [Train][Epoch 24/120][Avg]top1: 0.98211, CELoss: 0.04716, loss: 0.04716
还有一个问题，这个平均值是怎么算出来了，感觉怎么都不到0.98211 [2024/07/31 20:24:32] ppcls INFO: [Train][Epoch 24/120][Iter: 0/304]lr(PiecewiseDecay): 0.00100000, top1: 1.00000, CELoss: 0.01757, loss: 0.01757, batch_cost: 0.61280s, reader_cost: 0.12493, ips: 26.10981 samples/s, eta: 5:01:10 [2024/07/31 20:24:38] ppcls INFO: [Train][Epoch 24/120][Iter: 10/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97159, CELoss: 0.05098, loss: 0.05098, batch_cost: 0.62808s, reader_cost: 0.12837, ips: 25.47451 samples/s, eta: 5:08:34 [2024/07/31 20:24:44] ppcls INFO: [Train][Epoch 24/120][Iter: 20/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97619, CELoss: 0.05253, loss: 0.05253, batch_cost: 0.62088s, reader_cost: 0.12821, ips: 25.76972 samples/s, eta: 5:04:56 [2024/07/31 20:24:51] ppcls INFO: [Train][Epoch 24/120][Iter: 30/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97379, CELoss: 0.06434, loss: 0.06434, batch_cost: 0.62541s, reader_cost: 0.13184, ips: 25.58316 samples/s, eta: 5:07:03 [2024/07/31 20:24:57] ppcls INFO: [Train][Epoch 24/120][Iter: 40/304]lr(PiecewiseDecay): 0.00100000, top1: 0.96951, CELoss: 0.06668, loss: 0.06668, batch_cost: 0.62095s, reader_cost: 0.12972, ips: 25.76705 samples/s, eta: 5:04:45 [2024/07/31 20:25:03] ppcls INFO: [Train][Epoch 24/120][Iter: 50/304]lr(PiecewiseDecay): 0.00100000, top1: 0.96814, CELoss: 0.06468, loss: 0.06468, batch_cost: 0.61844s, reader_cost: 0.12773, ips: 25.87151 samples/s, eta: 5:03:25 [2024/07/31 20:25:09] ppcls INFO: [Train][Epoch 24/120][Iter: 60/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97029, CELoss: 0.06370, loss: 0.06370, batch_cost: 0.61698s, reader_cost: 0.12703, ips: 25.93262 samples/s, eta: 5:02:36 [2024/07/31 20:25:15] ppcls INFO: [Train][Epoch 24/120][Iter: 70/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97359, CELoss: 0.05873, loss: 0.05873, batch_cost: 0.61571s, reader_cost: 0.12661, ips: 25.98613 samples/s, eta: 5:01:53 [2024/07/31 20:25:21] ppcls INFO: [Train][Epoch 24/120][Iter: 80/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97531, CELoss: 0.05557, loss: 0.05557, batch_cost: 0.61537s, reader_cost: 0.12652, ips: 26.00050 samples/s, eta: 5:01:36 [2024/07/31 20:25:27] ppcls INFO: [Train][Epoch 24/120][Iter: 90/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97734, CELoss: 0.05107, loss: 0.05107, batch_cost: 0.61270s, reader_cost: 0.12442, ips: 26.11384 samples/s, eta: 5:00:12 [2024/07/31 20:25:33] ppcls INFO: [Train][Epoch 24/120][Iter: 100/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97772, CELoss: 0.05041, loss: 0.05041, batch_cost: 0.61239s, reader_cost: 0.12465, ips: 26.12733 samples/s, eta: 4:59:56 [2024/07/31 20:25:39] ppcls INFO: [Train][Epoch 24/120][Iter: 110/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97804, CELoss: 0.05052, loss: 0.05052, batch_cost: 0.61104s, reader_cost: 0.12344, ips: 26.18475 samples/s, eta: 4:59:11 [2024/07/31 20:25:45] ppcls INFO: [Train][Epoch 24/120][Iter: 120/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97882, CELoss: 0.04917, loss: 0.04917, batch_cost: 0.60958s, reader_cost: 0.12228, ips: 26.24745 samples/s, eta: 4:58:22 [2024/07/31 20:25:51] ppcls INFO: [Train][Epoch 24/120][Iter: 130/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97996, CELoss: 0.04716, loss: 0.04716, batch_cost: 0.60957s, reader_cost: 0.12241, ips: 26.24819 samples/s, eta: 4:58:15 [2024/07/31 20:25:57] ppcls INFO: [Train][Epoch 24/120][Iter: 140/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97961, CELoss: 0.05031, loss: 0.05031, batch_cost: 0.60854s, reader_cost: 0.12157, ips: 26.29241 samples/s, eta: 4:57:39 [2024/07/31 20:26:03] ppcls INFO: [Train][Epoch 24/120][Iter: 150/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97972, CELoss: 0.04904, loss: 0.04904, batch_cost: 0.60852s, reader_cost: 0.12143, ips: 26.29311 samples/s, eta: 4:57:32 [2024/07/31 20:26:10] ppcls INFO: [Train][Epoch 24/120][Iter: 160/304]lr(PiecewiseDecay): 0.00100000, top1: 0.98059, CELoss: 0.04801, loss: 0.04801, batch_cost: 0.60879s, reader_cost: 0.12156, ips: 26.28154 samples/s, eta: 4:57:34 [2024/07/31 20:26:15] ppcls INFO: [Train][Epoch 24/120][Iter: 170/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97990, CELoss: 0.04968, loss: 0.04968, batch_cost: 0.60820s, reader_cost: 0.12103, ips: 26.30717 samples/s, eta: 4:57:11 [2024/07/31 20:26:22] ppcls INFO: [Train][Epoch 24/120][Iter: 180/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97963, CELoss: 0.04977, loss: 0.04977, batch_cost: 0.60820s, reader_cost: 0.12117, ips: 26.30719 samples/s, eta: 4:57:05 [2024/07/31 20:26:28] ppcls INFO: [Train][Epoch 24/120][Iter: 190/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97971, CELoss: 0.04987, loss: 0.04987, batch_cost: 0.60803s, reader_cost: 0.12091, ips: 26.31468 samples/s, eta: 4:56:53 [2024/07/31 20:26:34] ppcls INFO: [Train][Epoch 24/120][Iter: 200/304]lr(PiecewiseDecay): 0.00100000, top1: 0.98010, CELoss: 0.05034, loss: 0.05034, batch_cost: 0.60766s, reader_cost: 0.12063, ips: 26.33051 samples/s, eta: 4:56:37 [2024/07/31 20:26:40] ppcls INFO: [Train][Epoch 24/120][Iter: 210/304]lr(PiecewiseDecay): 0.00100000, top1: 0.98015, CELoss: 0.05039, loss: 0.05039, batch_cost: 0.60726s, reader_cost: 0.12033, ips: 26.34805 samples/s, eta: 4:56:19 [2024/07/31 20:26:46] ppcls INFO: [Train][Epoch 24/120][Iter: 220/304]lr(PiecewiseDecay): 0.00100000, top1: 0.98077, CELoss: 0.04899, loss: 0.04899, batch_cost: 0.60762s, reader_cost: 0.12083, ips: 26.33223 samples/s, eta: 4:56:23 [2024/07/31 20:26:52] ppcls INFO: [Train][Epoch 24/120][Iter: 230/304]lr(PiecewiseDecay): 0.00100000, top1: 0.98079, CELoss: 0.04874, loss: 0.04874, batch_cost: 0.60726s, reader_cost: 0.12061, ips: 26.34782 samples/s, eta: 4:56:07 [2024/07/31 20:26:58] ppcls INFO: [Train][Epoch 24/120][Iter: 240/304]lr(PiecewiseDecay): 0.00100000, top1: 0.98107, CELoss: 0.04837, loss: 0.04837, batch_cost: 0.60674s, reader_cost: 0.12025, ips: 26.37032 samples/s, eta: 4:55:46 [2024/07/31 20:27:04] ppcls INFO: [Train][Epoch 24/120][Iter: 250/304]lr(PiecewiseDecay): 0.00100000, top1: 0.98157, CELoss: 0.04781, loss: 0.04781, batch_cost: 0.60659s, reader_cost: 0.12022, ips: 26.37714 samples/s, eta: 4:55:35 [2024/07/31 20:27:10] ppcls INFO: [Train][Epoch 24/120][Iter: 260/304]lr(PiecewiseDecay): 0.00100000, top1: 0.98156, CELoss: 0.04710, loss: 0.04710, batch_cost: 0.60626s, reader_cost: 0.12001, ips: 26.39147 samples/s, eta: 4:55:19 [2024/07/31 20:27:16] ppcls INFO: [Train][Epoch 24/120][Iter: 270/304]lr(PiecewiseDecay): 0.00100000, top1: 0.98155, CELoss: 0.04814, loss: 0.04814, batch_cost: 0.60576s, reader_cost: 0.11958, ips: 26.41314 samples/s, eta: 4:54:59 [2024/07/31 20:27:22] ppcls INFO: [Train][Epoch 24/120][Iter: 280/304]lr(PiecewiseDecay): 0.00100000, top1: 0.98132, CELoss: 0.04887, loss: 0.04887, batch_cost: 0.60638s, reader_cost: 0.12027, ips: 26.38607 samples/s, eta: 4:55:11 [2024/07/31 20:27:28] ppcls INFO: [Train][Epoch 24/120][Iter: 290/304]lr(PiecewiseDecay): 0.00100000, top1: 0.98174, CELoss: 0.04826, loss: 0.04826, batch_cost: 0.60602s, reader_cost: 0.11995, ips: 26.40185 samples/s, eta: 4:54:54 [2024/07/31 20:27:34] ppcls INFO: [Train][Epoch 24/120][Iter: 300/304]lr(PiecewiseDecay): 0.00100000, top1: 0.98214, CELoss: 0.04717, loss: 0.04717, batch_cost: 0.60588s, reader_cost: 0.11982, ips: 26.40767 samples/s, eta: 4:54:44 [2024/07/31 20:27:36] ppcls INFO: [Train][Epoch 24/120][Avg]top1: 0.98211, CELoss: 0.04716, loss: 0.04716
训练过程中的top1 acc 只针对当前批次模型预测结果，不是整个训练数据集的统计结果
在eval的时候没有加载训练好的权重
还有一个问题，这个平均值是怎么算出来了，感觉怎么都不到0.98211 [2024/07/31 20:24:32] ppcls INFO: [Train][Epoch 24/120][Iter: 0/304]lr(PiecewiseDecay): 0.00100000, top1: 1.00000, CELoss: 0.01757, loss: 0.01757, batch_cost: 0.61280s, reader_cost: 0.12493, ips: 26.10981 samples/s, eta: 5:01:10 [2024/07/31 20:24:38] ppcls INFO: [Train][Epoch 24/120][Iter: 10/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97159, CELoss: 0.05098, loss: 0.05098, batch_cost: 0.62808s, reader_cost: 0.12837, ips: 25.47451 samples/s, eta: 5:08:34 [2024/07/31 20:24:44] ppcls INFO: [Train][Epoch 24/120][Iter: 20/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97619, CELoss: 0.05253, loss: 0.05253, batch_cost: 0.62088s, reader_cost: 0.12821, ips: 25.76972 samples/s, eta: 5:04:56 [2024/07/31 20:24:51] ppcls INFO: [Train][Epoch 24/120][Iter: 30/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97379, CELoss: 0.06434, loss: 0.06434, batch_cost: 0.62541s, reader_cost: 0.13184, ips: 25.58316 samples/s, eta: 5:07:03 [2024/07/31 20:24:57] ppcls INFO: [Train][Epoch 24/120][Iter: 40/304]lr(PiecewiseDecay): 0.00100000, top1: 0.96951, CELoss: 0.06668, loss: 0.06668, batch_cost: 0.62095s, reader_cost: 0.12972, ips: 25.76705 samples/s, eta: 5:04:45 [2024/07/31 20:25:03] ppcls INFO: [Train][Epoch 24/120][Iter: 50/304]lr(PiecewiseDecay): 0.00100000, top1: 0.96814, CELoss: 0.06468, loss: 0.06468, batch_cost: 0.61844s, reader_cost: 0.12773, ips: 25.87151 samples/s, eta: 5:03:25 [2024/07/31 20:25:09] ppcls INFO: [Train][Epoch 24/120][Iter: 60/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97029, CELoss: 0.06370, loss: 0.06370, batch_cost: 0.61698s, reader_cost: 0.12703, ips: 25.93262 samples/s, eta: 5:02:36 [2024/07/31 20:25:15] ppcls INFO: [Train][Epoch 24/120][Iter: 70/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97359, CELoss: 0.05873, loss: 0.05873, batch_cost: 0.61571s, reader_cost: 0.12661, ips: 25.98613 samples/s, eta: 5:01:53 [2024/07/31 20:25:21] ppcls INFO: [Train][Epoch 24/120][Iter: 80/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97531, CELoss: 0.05557, loss: 0.05557, batch_cost: 0.61537s, reader_cost: 0.12652, ips: 26.00050 samples/s, eta: 5:01:36 [2024/07/31 20:25:27] ppcls INFO: [Train][Epoch 24/120][Iter: 90/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97734, CELoss: 0.05107, loss: 0.05107, batch_cost: 0.61270s, reader_cost: 0.12442, ips: 26.11384 samples/s, eta: 5:00:12 [2024/07/31 20:25:33] ppcls INFO: [Train][Epoch 24/120][Iter: 100/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97772, CELoss: 0.05041, loss: 0.05041, batch_cost: 0.61239s, reader_cost: 0.12465, ips: 26.12733 samples/s, eta: 4:59:56 [2024/07/31 20:25:39] ppcls INFO: [Train][Epoch 24/120][Iter: 110/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97804, CELoss: 0.05052, loss: 0.05052, batch_cost: 0.61104s, reader_cost: 0.12344, ips: 26.18475 samples/s, eta: 4:59:11 [2024/07/31 20:25:45] ppcls INFO: [Train][Epoch 24/120][Iter: 120/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97882, CELoss: 0.04917, loss: 0.04917, batch_cost: 0.60958s, reader_cost: 0.12228, ips: 26.24745 samples/s, eta: 4:58:22 [2024/07/31 20:25:51] ppcls INFO: [Train][Epoch 24/120][Iter: 130/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97996, CELoss: 0.04716, loss: 0.04716, batch_cost: 0.60957s, reader_cost: 0.12241, ips: 26.24819 samples/s, eta: 4:58:15 [2024/07/31 20:25:57] ppcls INFO: [Train][Epoch 24/120][Iter: 140/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97961, CELoss: 0.05031, loss: 0.05031, batch_cost: 0.60854s, reader_cost: 0.12157, ips: 26.29241 samples/s, eta: 4:57:39 [2024/07/31 20:26:03] ppcls INFO: [Train][Epoch 24/120][Iter: 150/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97972, CELoss: 0.04904, loss: 0.04904, batch_cost: 0.60852s, reader_cost: 0.12143, ips: 26.29311 samples/s, eta: 4:57:32 [2024/07/31 20:26:10] ppcls INFO: [Train][Epoch 24/120][Iter: 160/304]lr(PiecewiseDecay): 0.00100000, top1: 0.98059, CELoss: 0.04801, loss: 0.04801, batch_cost: 0.60879s, reader_cost: 0.12156, ips: 26.28154 samples/s, eta: 4:57:34 [2024/07/31 20:26:15] ppcls INFO: [Train][Epoch 24/120][Iter: 170/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97990, CELoss: 0.04968, loss: 0.04968, batch_cost: 0.60820s, reader_cost: 0.12103, ips: 26.30717 samples/s, eta: 4:57:11 [2024/07/31 20:26:22] ppcls INFO: [Train][Epoch 24/120][Iter: 180/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97963, CELoss: 0.04977, loss: 0.04977, batch_cost: 0.60820s, reader_cost: 0.12117, ips: 26.30719 samples/s, eta: 4:57:05 [2024/07/31 20:26:28] ppcls INFO: [Train][Epoch 24/120][Iter: 190/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97971, CELoss: 0.04987, loss: 0.04987, batch_cost: 0.60803s, reader_cost: 0.12091, ips: 26.31468 samples/s, eta: 4:56:53 [2024/07/31 20:26:34] ppcls INFO: [Train][Epoch 24/120][Iter: 200/304]lr(PiecewiseDecay): 0.00100000, top1: 0.98010, CELoss: 0.05034, loss: 0.05034, batch_cost: 0.60766s, reader_cost: 0.12063, ips: 26.33051 samples/s, eta: 4:56:37 [2024/07/31 20:26:40] ppcls INFO: [Train][Epoch 24/120][Iter: 210/304]lr(PiecewiseDecay): 0.00100000, top1: 0.98015, CELoss: 0.05039, loss: 0.05039, batch_cost: 0.60726s, reader_cost: 0.12033, ips: 26.34805 samples/s, eta: 4:56:19 [2024/07/31 20:26:46] ppcls INFO: [Train][Epoch 24/120][Iter: 220/304]lr(PiecewiseDecay): 0.00100000, top1: 0.98077, CELoss: 0.04899, loss: 0.04899, batch_cost: 0.60762s, reader_cost: 0.12083, ips: 26.33223 samples/s, eta: 4:56:23 [2024/07/31 20:26:52] ppcls INFO: [Train][Epoch 24/120][Iter: 230/304]lr(PiecewiseDecay): 0.00100000, top1: 0.98079, CELoss: 0.04874, loss: 0.04874, batch_cost: 0.60726s, reader_cost: 0.12061, ips: 26.34782 samples/s, eta: 4:56:07 [2024/07/31 20:26:58] ppcls INFO: [Train][Epoch 24/120][Iter: 240/304]lr(PiecewiseDecay): 0.00100000, top1: 0.98107, CELoss: 0.04837, loss: 0.04837, batch_cost: 0.60674s, reader_cost: 0.12025, ips: 26.37032 samples/s, eta: 4:55:46 [2024/07/31 20:27:04] ppcls INFO: [Train][Epoch 24/120][Iter: 250/304]lr(PiecewiseDecay): 0.00100000, top1: 0.98157, CELoss: 0.04781, loss: 0.04781, batch_cost: 0.60659s, reader_cost: 0.12022, ips: 26.37714 samples/s, eta: 4:55:35 [2024/07/31 20:27:10] ppcls INFO: [Train][Epoch 24/120][Iter: 260/304]lr(PiecewiseDecay): 0.00100000, top1: 0.98156, CELoss: 0.04710, loss: 0.04710, batch_cost: 0.60626s, reader_cost: 0.12001, ips: 26.39147 samples/s, eta: 4:55:19 [2024/07/31 20:27:16] ppcls INFO: [Train][Epoch 24/120][Iter: 270/304]lr(PiecewiseDecay): 0.00100000, top1: 0.98155, CELoss: 0.04814, loss: 0.04814, batch_cost: 0.60576s, reader_cost: 0.11958, ips: 26.41314 samples/s, eta: 4:54:59 [2024/07/31 20:27:22] ppcls INFO: [Train][Epoch 24/120][Iter: 280/304]lr(PiecewiseDecay): 0.00100000, top1: 0.98132, CELoss: 0.04887, loss: 0.04887, batch_cost: 0.60638s, reader_cost: 0.12027, ips: 26.38607 samples/s, eta: 4:55:11 [2024/07/31 20:27:28] ppcls INFO: [Train][Epoch 24/120][Iter: 290/304]lr(PiecewiseDecay): 0.00100000, top1: 0.98174, CELoss: 0.04826, loss: 0.04826, batch_cost: 0.60602s, reader_cost: 0.11995, ips: 26.40185 samples/s, eta: 4:54:54 [2024/07/31 20:27:34] ppcls INFO: [Train][Epoch 24/120][Iter: 300/304]lr(PiecewiseDecay): 0.00100000, top1: 0.98214, CELoss: 0.04717, loss: 0.04717, batch_cost: 0.60588s, reader_cost: 0.11982, ips: 26.40767 samples/s, eta: 4:54:44 [2024/07/31 20:27:36] ppcls INFO: [Train][Epoch 24/120][Avg]top1: 0.98211, CELoss: 0.04716, loss: 0.04716
训练过程中的top1 acc 只针对当前批次模型预测结果，不是整个训练数据集的统计结果
咱们选最佳模型的依据是什么？我们是否可以信任选择的最佳模型？
还有一个问题，这个平均值是怎么算出来了，感觉怎么都不到0.98211 [2024/07/31 20:24:32] ppcls INFO: [Train][Epoch 24/120][Iter: 0/304]lr(PiecewiseDecay): 0.00100000, top1: 1.00000, CELoss: 0.01757, loss: 0.01757, batch_cost: 0.61280s, reader_cost: 0.12493, ips: 26.10981 samples/s, eta: 5:01:10 [2024/07/31 20:24:38] ppcls INFO: [Train][Epoch 24/120][Iter: 10/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97159, CELoss: 0.05098, loss: 0.05098, batch_cost: 0.62808s, reader_cost: 0.12837, ips: 25.47451 samples/s, eta: 5:08:34 [2024/07/31 20:24:44] ppcls INFO: [Train][Epoch 24/120][Iter: 20/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97619, CELoss: 0.05253, loss: 0.05253, batch_cost: 0.62088s, reader_cost: 0.12821, ips: 25.76972 samples/s, eta: 5:04:56 [2024/07/31 20:24:51] ppcls INFO: [Train][Epoch 24/120][Iter: 30/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97379, CELoss: 0.06434, loss: 0.06434, batch_cost: 0.62541s, reader_cost: 0.13184, ips: 25.58316 samples/s, eta: 5:07:03 [2024/07/31 20:24:57] ppcls INFO: [Train][Epoch 24/120][Iter: 40/304]lr(PiecewiseDecay): 0.00100000, top1: 0.96951, CELoss: 0.06668, loss: 0.06668, batch_cost: 0.62095s, reader_cost: 0.12972, ips: 25.76705 samples/s, eta: 5:04:45 [2024/07/31 20:25:03] ppcls INFO: [Train][Epoch 24/120][Iter: 50/304]lr(PiecewiseDecay): 0.00100000, top1: 0.96814, CELoss: 0.06468, loss: 0.06468, batch_cost: 0.61844s, reader_cost: 0.12773, ips: 25.87151 samples/s, eta: 5:03:25 [2024/07/31 20:25:09] ppcls INFO: [Train][Epoch 24/120][Iter: 60/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97029, CELoss: 0.06370, loss: 0.06370, batch_cost: 0.61698s, reader_cost: 0.12703, ips: 25.93262 samples/s, eta: 5:02:36 [2024/07/31 20:25:15] ppcls INFO: [Train][Epoch 24/120][Iter: 70/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97359, CELoss: 0.05873, loss: 0.05873, batch_cost: 0.61571s, reader_cost: 0.12661, ips: 25.98613 samples/s, eta: 5:01:53 [2024/07/31 20:25:21] ppcls INFO: [Train][Epoch 24/120][Iter: 80/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97531, CELoss: 0.05557, loss: 0.05557, batch_cost: 0.61537s, reader_cost: 0.12652, ips: 26.00050 samples/s, eta: 5:01:36 [2024/07/31 20:25:27] ppcls INFO: [Train][Epoch 24/120][Iter: 90/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97734, CELoss: 0.05107, loss: 0.05107, batch_cost: 0.61270s, reader_cost: 0.12442, ips: 26.11384 samples/s, eta: 5:00:12 [2024/07/31 20:25:33] ppcls INFO: [Train][Epoch 24/120][Iter: 100/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97772, CELoss: 0.05041, loss: 0.05041, batch_cost: 0.61239s, reader_cost: 0.12465, ips: 26.12733 samples/s, eta: 4:59:56 [2024/07/31 20:25:39] ppcls INFO: [Train][Epoch 24/120][Iter: 110/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97804, CELoss: 0.05052, loss: 0.05052, batch_cost: 0.61104s, reader_cost: 0.12344, ips: 26.18475 samples/s, eta: 4:59:11 [2024/07/31 20:25:45] ppcls INFO: [Train][Epoch 24/120][Iter: 120/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97882, CELoss: 0.04917, loss: 0.04917, batch_cost: 0.60958s, reader_cost: 0.12228, ips: 26.24745 samples/s, eta: 4:58:22 [2024/07/31 20:25:51] ppcls INFO: [Train][Epoch 24/120][Iter: 130/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97996, CELoss: 0.04716, loss: 0.04716, batch_cost: 0.60957s, reader_cost: 0.12241, ips: 26.24819 samples/s, eta: 4:58:15 [2024/07/31 20:25:57] ppcls INFO: [Train][Epoch 24/120][Iter: 140/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97961, CELoss: 0.05031, loss: 0.05031, batch_cost: 0.60854s, reader_cost: 0.12157, ips: 26.29241 samples/s, eta: 4:57:39 [2024/07/31 20:26:03] ppcls INFO: [Train][Epoch 24/120][Iter: 150/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97972, CELoss: 0.04904, loss: 0.04904, batch_cost: 0.60852s, reader_cost: 0.12143, ips: 26.29311 samples/s, eta: 4:57:32 [2024/07/31 20:26:10] ppcls INFO: [Train][Epoch 24/120][Iter: 160/304]lr(PiecewiseDecay): 0.00100000, top1: 0.98059, CELoss: 0.04801, loss: 0.04801, batch_cost: 0.60879s, reader_cost: 0.12156, ips: 26.28154 samples/s, eta: 4:57:34 [2024/07/31 20:26:15] ppcls INFO: [Train][Epoch 24/120][Iter: 170/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97990, CELoss: 0.04968, loss: 0.04968, batch_cost: 0.60820s, reader_cost: 0.12103, ips: 26.30717 samples/s, eta: 4:57:11 [2024/07/31 20:26:22] ppcls INFO: [Train][Epoch 24/120][Iter: 180/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97963, CELoss: 0.04977, loss: 0.04977, batch_cost: 0.60820s, reader_cost: 0.12117, ips: 26.30719 samples/s, eta: 4:57:05 [2024/07/31 20:26:28] ppcls INFO: [Train][Epoch 24/120][Iter: 190/304]lr(PiecewiseDecay): 0.00100000, top1: 0.97971, CELoss: 0.04987, loss: 0.04987, batch_cost: 0.60803s, reader_cost: 0.12091, ips: 26.31468 samples/s, eta: 4:56:53 [2024/07/31 20:26:34] ppcls INFO: [Train][Epoch 24/120][Iter: 200/304]lr(PiecewiseDecay): 0.00100000, top1: 0.98010, CELoss: 0.05034, loss: 0.05034, batch_cost: 0.60766s, reader_cost: 0.12063, ips: 26.33051 samples/s, eta: 4:56:37 [2024/07/31 20:26:40] ppcls INFO: [Train][Epoch 24/120][Iter: 210/304]lr(PiecewiseDecay): 0.00100000, top1: 0.98015, CELoss: 0.05039, loss: 0.05039, batch_cost: 0.60726s, reader_cost: 0.12033, ips: 26.34805 samples/s, eta: 4:56:19 [2024/07/31 20:26:46] ppcls INFO: [Train][Epoch 24/120][Iter: 220/304]lr(PiecewiseDecay): 0.00100000, top1: 0.98077, CELoss: 0.04899, loss: 0.04899, batch_cost: 0.60762s, reader_cost: 0.12083, ips: 26.33223 samples/s, eta: 4:56:23 [2024/07/31 20:26:52] ppcls INFO: [Train][Epoch 24/120][Iter: 230/304]lr(PiecewiseDecay): 0.00100000, top1: 0.98079, CELoss: 0.04874, loss: 0.04874, batch_cost: 0.60726s, reader_cost: 0.12061, ips: 26.34782 samples/s, eta: 4:56:07 [2024/07/31 20:26:58] ppcls INFO: [Train][Epoch 24/120][Iter: 240/304]lr(PiecewiseDecay): 0.00100000, top1: 0.98107, CELoss: 0.04837, loss: 0.04837, batch_cost: 0.60674s, reader_cost: 0.12025, ips: 26.37032 samples/s, eta: 4:55:46 [2024/07/31 20:27:04] ppcls INFO: [Train][Epoch 24/120][Iter: 250/304]lr(PiecewiseDecay): 0.00100000, top1: 0.98157, CELoss: 0.04781, loss: 0.04781, batch_cost: 0.60659s, reader_cost: 0.12022, ips: 26.37714 samples/s, eta: 4:55:35 [2024/07/31 20:27:10] ppcls INFO: [Train][Epoch 24/120][Iter: 260/304]lr(PiecewiseDecay): 0.00100000, top1: 0.98156, CELoss: 0.04710, loss: 0.04710, batch_cost: 0.60626s, reader_cost: 0.12001, ips: 26.39147 samples/s, eta: 4:55:19 [2024/07/31 20:27:16] ppcls INFO: [Train][Epoch 24/120][Iter: 270/304]lr(PiecewiseDecay): 0.00100000, top1: 0.98155, CELoss: 0.04814, loss: 0.04814, batch_cost: 0.60576s, reader_cost: 0.11958, ips: 26.41314 samples/s, eta: 4:54:59 [2024/07/31 20:27:22] ppcls INFO: [Train][Epoch 24/120][Iter: 280/304]lr(PiecewiseDecay): 0.00100000, top1: 0.98132, CELoss: 0.04887, loss: 0.04887, batch_cost: 0.60638s, reader_cost: 0.12027, ips: 26.38607 samples/s, eta: 4:55:11 [2024/07/31 20:27:28] ppcls INFO: [Train][Epoch 24/120][Iter: 290/304]lr(PiecewiseDecay): 0.00100000, top1: 0.98174, CELoss: 0.04826, loss: 0.04826, batch_cost: 0.60602s, reader_cost: 0.11995, ips: 26.40185 samples/s, eta: 4:54:54 [2024/07/31 20:27:34] ppcls INFO: [Train][Epoch 24/120][Iter: 300/304]lr(PiecewiseDecay): 0.00100000, top1: 0.98214, CELoss: 0.04717, loss: 0.04717, batch_cost: 0.60588s, reader_cost: 0.11982, ips: 26.40767 samples/s, eta: 4:54:44 [2024/07/31 20:27:36] ppcls INFO: [Train][Epoch 24/120][Avg]top1: 0.98211, CELoss: 0.04716, loss: 0.04716
训练过程中的top1 acc 只针对当前批次模型预测结果，不是整个训练数据集的统计结果
咱们选最佳模型的依据是什么？我们是否可以信任选择的最佳模型？
这模型是你训练出来的权重，在输出路径下有个best_model.pdparams,在评估的时候设置 -o Global.pretrained_model=xxx/best_model.pdparams
The issue has no response for a long time and will be closed. You can reopen or new another issue if are still confused.
From Bot
The issue has no response for a long time and will be closed. You can reopen or new another issue if are still confused.
From Bot