From 6083e9ce4abd1c11429eb64ab7fa7bf363080ac9 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=83=91=E7=89=B9=E9=A9=B9?= Date: Tue, 13 May 2025 16:34:18 +0800 Subject: [PATCH] =?UTF-8?q?=E6=9B=B4=E6=96=B0=E9=9A=8F=E7=89=88=E6=9C=AC?= =?UTF-8?q?=E6=BC=94=E8=BF=9B=E6=A8=A1=E5=9E=8B=E7=9A=84=E6=80=A7=E8=83=BD?= =?UTF-8?q?=E6=95=B0=E6=8D=AE?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- PyTorch/README.CN.md | 1 + PyTorch/built-in/audio/ESPnet2_for_PyTorch/README.md | 4 ++-- .../built-in/audio/Wenet_Conformer_for_Pytorch/README.md | 6 +++--- .../built-in/cv/classification/CRNN_for_PyTorch/README.md | 4 ++-- .../cv/classification/EfficientNetV2_for_PyTorch/README.md | 4 ++-- .../MobileNetV3-Large_ID1784_for_PyTorch/README.md | 4 ++-- .../cv/classification/ResNet50_ID4149_for_PyTorch/README.md | 4 ++-- .../cv/classification/Resnet50_Cifar_for_PyTorch/README.md | 4 ++-- .../built-in/cv/detection/DB_ID0706_for_PyTorch/README.md | 4 ++-- .../cv/detection/Faster_Mask_RCNN_for_PyTorch/README.md | 4 ++-- PyTorch/built-in/mm/CLIP_for_PyTorch/README.md | 5 +++++ .../built-in/nlp/Bert-Squad_ID0470_for_PyTorch/README.md | 6 +++--- .../built-in/nlp/Bert_Chinese_ID3433_for_PyTorch/README.md | 4 ++-- .../nlp/Fairseq_Transformer_wmt18_for_PyTorch/README.md | 4 ++-- PyTorch/built-in/rl/MAPPO_for_PyTorch/README.md | 4 ++-- PyTorch/built-in/rl/PPO_for_Pytorch/README.md | 4 ++-- PyTorch/contrib/audio/wav2vec2.0/README.md | 4 ++-- .../cv/classification/HRNet_ID1780_for_PyTorch/README.md | 4 ++-- .../classification/InceptionV3_ID1596_for_PyTorch/README.md | 4 ++-- PyTorch/contrib/cv/classification/MAE_for_PyTorch/README.md | 4 ++-- .../cv/pose_estimation/Lightweight_OpenPose/README.md | 4 ++-- .../semantic_segmentation/DeeplabV3_for_Pytorch/README.md | 4 ++-- PyTorch/contrib/nlp/BERT-NER-Pytorch/README.md | 4 ++-- PyTorch/contrib/nlp/albert_ID0335_for_PyTorch/README.md | 4 ++-- PyTorch/contrib/nlp/roberta_for_PyTorch/README.md | 4 ++-- PyTorch/dev/perf/ShuffleNetV2_iflytek_for_Pytorch/README.md | 4 ++-- 26 files changed, 56 insertions(+), 50 deletions(-) diff --git a/PyTorch/README.CN.md b/PyTorch/README.CN.md index 1af45455bd..f36f5b5199 100644 --- a/PyTorch/README.CN.md +++ b/PyTorch/README.CN.md @@ -72,6 +72,7 @@ ### 随版本演进模型 +随版本演进模型的最新性能数据均在OpenEuler操作系统下实测获得。如您使用的是其他操作系统,则展示的数据仅作为参考。 - [albert_ID0335_for_PyTorch(FP16)](https://gitee.com/ascend/ModelZoo-PyTorch/tree/master/PyTorch/contrib/nlp/albert_ID0335_for_PyTorch) - [Bert_Chinese_ID3433_for_PyTorch(FP16)](https://gitee.com/ascend/ModelZoo-PyTorch/tree/master/PyTorch/built-in/nlp/Bert_Chinese_ID3433_for_PyTorch) - [BERT-NER-Pytorch(FP32)](https://gitee.com/ascend/ModelZoo-PyTorch/tree/master/PyTorch/contrib/nlp/BERT-NER-Pytorch) diff --git a/PyTorch/built-in/audio/ESPnet2_for_PyTorch/README.md b/PyTorch/built-in/audio/ESPnet2_for_PyTorch/README.md index 25e2a26594..c82435c8a3 100644 --- a/PyTorch/built-in/audio/ESPnet2_for_PyTorch/README.md +++ b/PyTorch/built-in/audio/ESPnet2_for_PyTorch/README.md @@ -206,11 +206,11 @@ ESPNet是一套基于E2E的开源工具包,可进行语音识别等任务。 | 8p-NPU | 混合精度 | 95.4 | 751.37 | 50 | 1.11 | | 8p-NPU | 混合精度 | 95.4 | 700.96 | 50 | 2.1 | -说明:上表为历史数据,仅供参考。2024年12月31日更新的性能数据如下: +说明:上表为历史数据,仅供参考。2025年5月10日更新的性能数据如下: | NAME | 精度类型 | FPS | | :------ |:-------:|:------:| | 8p-竞品 | FP16 | 700.96 | -| 8p-Atlas 900 A2 PoDc | FP16 | 933.35 | +| 8p-Atlas 900 A2 PoDc | FP16 | 765.56 | # 版本说明 diff --git a/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/README.md b/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/README.md index c611e870b8..d56aab9539 100644 --- a/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/README.md +++ b/PyTorch/built-in/audio/Wenet_Conformer_for_Pytorch/README.md @@ -197,13 +197,13 @@ Wenet是一款开源的、面向工业落地应用的语音识别工具包,主 |:------:|:----:| | ctc greedy search | 4.96 | -说明:上表为历史数据,仅供参考。2024年12月31日更新的性能数据如下: +说明:上表为历史数据,仅供参考。2025年5月10日更新的性能数据如下: | NAME | 精度类型 | FPS | | :------ |:-------:|:------:| | Conformer 8p-竞品 | FP32 | 958.98 | -| Conformer 8p-Atlas 900 A2 PoDc | FP32 | 1435.9 | +| Conformer 8p-Atlas 900 A2 PoDc | FP32 | 1166.96 | | whisper 8p-竞品 | FP16 | 748.85 | -| whisper 8p-Atlas 900 A2 PoDc | FP16 | 1936.71 | +| whisper 8p-Atlas 900 A2 PoDc | FP16 | 1532.5 | # 公网地址说明 diff --git a/PyTorch/built-in/cv/classification/CRNN_for_PyTorch/README.md b/PyTorch/built-in/cv/classification/CRNN_for_PyTorch/README.md index d8455e49bd..2dbcfa49fd 100644 --- a/PyTorch/built-in/cv/classification/CRNN_for_PyTorch/README.md +++ b/PyTorch/built-in/cv/classification/CRNN_for_PyTorch/README.md @@ -176,11 +176,11 @@ CRNN (Convolutional Recurrent Neural Network) 于2015年由华中科技大学的 | 1p-NPU | - | 14078.58 | 1 | O2 | 2.1 | | 8p-NPU | 0.75 | 110879.797 | 100 | O2 | 2.1 | -说明:上表为历史数据,仅供参考。2024年12月31日更新的性能数据如下: +说明:上表为历史数据,仅供参考。2025年5月10日更新的性能数据如下: | NAME | 精度类型 | FPS | | :------ |:-------:|:------:| | 8p-竞品 | FP16 | 113122.50 | -| 8p-Atlas 900 A2 PoDc | FP16 | 127695.13 | +| 8p-Atlas 900 A2 PoDc | FP16 | 125304.48 | # 版本说明 2022.02.17:更新readme,重新发布。 diff --git a/PyTorch/built-in/cv/classification/EfficientNetV2_for_PyTorch/README.md b/PyTorch/built-in/cv/classification/EfficientNetV2_for_PyTorch/README.md index 5b51015879..749e51609b 100644 --- a/PyTorch/built-in/cv/classification/EfficientNetV2_for_PyTorch/README.md +++ b/PyTorch/built-in/cv/classification/EfficientNetV2_for_PyTorch/README.md @@ -195,11 +195,11 @@ EfficientNetV2是Efficient的改进版,accuracy达到了发布时的SOTA水平 | 1p-NPU | - | 1100.73 | 1 | O1 | 2.1 | | 8p-NPU | 82.19 | 6914.44 | 350 | O1 | 2.1 | -说明:上表为历史数据,仅供参考。2024年12月31日更新的性能数据如下: +说明:上表为历史数据,仅供参考。2025年5月10日更新的性能数据如下: | NAME | 精度类型 | FPS | | :------ |:-------:|:------:| | 8p-竞品 | FP16 | 6914.44 | -| 8p-Atlas 900 A2 PoDc | FP16 | 9273.32 | +| 8p-Atlas 900 A2 PoDc | FP16 | 7603.85 | # 版本说明 diff --git a/PyTorch/built-in/cv/classification/MobileNetV3-Large_ID1784_for_PyTorch/README.md b/PyTorch/built-in/cv/classification/MobileNetV3-Large_ID1784_for_PyTorch/README.md index c2f6f52dbd..81245d94e0 100644 --- a/PyTorch/built-in/cv/classification/MobileNetV3-Large_ID1784_for_PyTorch/README.md +++ b/PyTorch/built-in/cv/classification/MobileNetV3-Large_ID1784_for_PyTorch/README.md @@ -199,11 +199,11 @@ MobileNetV3结合了MobileNetV1的深度可分离卷积、MobileNetV2的Inverted | 1p-NPU | - | 1825.02 | 1 | O2 | 2.1 | | 8p-NPU | 73.5 | 9451.71 | 600 | O2 | 2.1 | -说明:上表为历史数据,仅供参考。2024年12月31日更新的性能数据如下: +说明:上表为历史数据,仅供参考。2025年5月10日更新的性能数据如下: | NAME | 精度类型 | FPS | | :------ |:-------:|:------:| | 8p-竞品 | FP16 | 9451.71 | -| 8p-Atlas 900 A2 PoDc | FP16 | 14462.5 | +| 8p-Atlas 900 A2 PoDc | FP16 | 12524 | # 版本说明 diff --git a/PyTorch/built-in/cv/classification/ResNet50_ID4149_for_PyTorch/README.md b/PyTorch/built-in/cv/classification/ResNet50_ID4149_for_PyTorch/README.md index cdaf12f060..f4ad614064 100644 --- a/PyTorch/built-in/cv/classification/ResNet50_ID4149_for_PyTorch/README.md +++ b/PyTorch/built-in/cv/classification/ResNet50_ID4149_for_PyTorch/README.md @@ -228,11 +228,11 @@ ResNet是由微软研究院的Kaiming He等四名华人提出,是ImageNet竞 | 8p-NPU | 76.3 | 13255 | 90 | O2 | 2.1 | | 16p-NPU | 76.69 | 30000 | 90 | O2 | 2.1 | -说明:上表为历史数据,仅供参考。2024年12月31日更新的性能数据如下: +说明:上表为历史数据,仅供参考。2025年5月10日更新的性能数据如下: | NAME | 精度类型 | FPS | | :------ |:-------:|:------:| | 8p-竞品 | FP16 | 13255 | -| 8p-Atlas 900 A2 PoDc | FP16 | 14321 | +| 8p-Atlas 900 A2 PoDc | FP16 | 14371 | # 版本说明 diff --git a/PyTorch/built-in/cv/classification/Resnet50_Cifar_for_PyTorch/README.md b/PyTorch/built-in/cv/classification/Resnet50_Cifar_for_PyTorch/README.md index 60e2c99f8e..1ccd75283c 100644 --- a/PyTorch/built-in/cv/classification/Resnet50_Cifar_for_PyTorch/README.md +++ b/PyTorch/built-in/cv/classification/Resnet50_Cifar_for_PyTorch/README.md @@ -247,11 +247,11 @@ MMClassification 是一款基于 PyTorch 的开源图像分类工具箱,是 Op | 1p-vNPU-X86 | 88.450 | 969 | 90 | | 2.1 | 16 | Atlas 200T A2 Box16 | 同等超参下,vNPU能满足精度要求 -说明:上表为历史数据,仅供参考。2024年12月31日更新的性能数据如下: +说明:上表为历史数据,仅供参考。2025年5月10日更新的性能数据如下: | NAME | 精度类型 | FPS | | :------ |:-------:|:------:| | 8p-竞品 | FP16 | 33032 | -| 8p-Atlas 900 A2 PoDc | FP16 | 38641 | +| 8p-Atlas 900 A2 PoDc | FP16 | 36571 | # 版本说明 diff --git a/PyTorch/built-in/cv/detection/DB_ID0706_for_PyTorch/README.md b/PyTorch/built-in/cv/detection/DB_ID0706_for_PyTorch/README.md index 809446734f..43515f60b2 100644 --- a/PyTorch/built-in/cv/detection/DB_ID0706_for_PyTorch/README.md +++ b/PyTorch/built-in/cv/detection/DB_ID0706_for_PyTorch/README.md @@ -204,11 +204,11 @@ DB(Differentiable Binarization)是一种使用可微分二值图来实时文字 | 1p-NPU | - | 29.926 | 1 | O2 | 2.1 | | 8p-NPU | 0.907 | 205.123 | 1200 | O2 | 2.1 | -说明:上表为历史数据,仅供参考。2024年12月31日更新的性能数据如下: +说明:上表为历史数据,仅供参考。2025年5月10日更新的性能数据如下: | NAME | 精度类型 | FPS | | :------ |:-------:|:------:| | 8p-竞品 | FP16 | 179.36 | -| 8p-Atlas 900 A2 PoDc | FP16 | 241.98 | +| 8p-Atlas 900 A2 PoDc | FP16 | 203.4 | # 版本说明 diff --git a/PyTorch/built-in/cv/detection/Faster_Mask_RCNN_for_PyTorch/README.md b/PyTorch/built-in/cv/detection/Faster_Mask_RCNN_for_PyTorch/README.md index 1a68bd6634..a7909466a2 100644 --- a/PyTorch/built-in/cv/detection/Faster_Mask_RCNN_for_PyTorch/README.md +++ b/PyTorch/built-in/cv/detection/Faster_Mask_RCNN_for_PyTorch/README.md @@ -297,11 +297,11 @@ vNPU环境下faster_rcnn结果比对 | 1p-vNPU- X86 | 0.326 | 11.075 | 90000 | | 2.1 | 8 | Atlas 200T A2 Box16 | 同等超参下,vNPU能满足精度要求 -说明:上表为历史数据,仅供参考。2024年12月31日更新的性能数据如下: +说明:上表为历史数据,仅供参考。2025年5月10日更新的性能数据如下: | NAME | 精度类型 | FPS | | :------ |:-------:|:------:| | 8p-竞品 | HF32 | 53.685 | -| 8p-Atlas 900 A2 PoDc | HF32 | 44.1 | +| 8p-Atlas 900 A2 PoDc | HF32 | 42.79 | # 版本说明 diff --git a/PyTorch/built-in/mm/CLIP_for_PyTorch/README.md b/PyTorch/built-in/mm/CLIP_for_PyTorch/README.md index 892d5eb462..a5a3a1b0b4 100644 --- a/PyTorch/built-in/mm/CLIP_for_PyTorch/README.md +++ b/PyTorch/built-in/mm/CLIP_for_PyTorch/README.md @@ -221,6 +221,11 @@ CLIP (Contrastive Language-Image Pre-Training,以下简称 CLIP) 模型是 Ope | 1p-NPU | - | 396.208 | O2 | 3 | 64 | 2.1 | | 8p-NPU | 1.5812 | 2924.63 | O2 | 3 | 64 | 2.1 | +说明:上表为历史数据,仅供参考。2025年5月10日更新的性能数据如下: +| NAME | 精度类型 | FPS | +| :------ |:-------:|:------:| +| CLIP 8p-竞品 | FP32 | 2924.63 | +| CLIP 8p-Atlas 900 A2 PoDc | FP32 | 3116.2 | # 版本说明 diff --git a/PyTorch/built-in/nlp/Bert-Squad_ID0470_for_PyTorch/README.md b/PyTorch/built-in/nlp/Bert-Squad_ID0470_for_PyTorch/README.md index b9ba1bcd0d..6e8657b385 100644 --- a/PyTorch/built-in/nlp/Bert-Squad_ID0470_for_PyTorch/README.md +++ b/PyTorch/built-in/nlp/Bert-Squad_ID0470_for_PyTorch/README.md @@ -209,13 +209,13 @@ BERT-Large模型是一个24层,1024维,24个自注意头(self attention he | 1p-bert-large | - | 106 | 1 | O2 | 2.1 | | 8p-bert-large | 90.87 | 829 | 833 | O2 | 2.1 | -说明:上表为历史数据,仅供参考。2024年12月31日更新的性能数据如下: +说明:上表为历史数据,仅供参考。2025年5月10日更新的性能数据如下: | NAME | 精度类型 | FPS | | :------ |:-------:|:------:| | Bert-Base 8p-竞品 | FP16 | 2463 | -| Bert-Base 8p-Atlas 900 A2 PoDc | FP16 | 3032 | +| Bert-Base 8p-Atlas 900 A2 PoDc | FP16 | 3118 | | Bert-Large 8p-竞品 | FP16 | 870.4 | -| Bert-Large 8p-Atlas 900 A2 PoDc | FP16 | 1077 | +| Bert-Large 8p-Atlas 900 A2 PoDc | FP16 | 1084 | # 公网地址说明 diff --git a/PyTorch/built-in/nlp/Bert_Chinese_ID3433_for_PyTorch/README.md b/PyTorch/built-in/nlp/Bert_Chinese_ID3433_for_PyTorch/README.md index 40e96ad44a..2fefdeff03 100644 --- a/PyTorch/built-in/nlp/Bert_Chinese_ID3433_for_PyTorch/README.md +++ b/PyTorch/built-in/nlp/Bert_Chinese_ID3433_for_PyTorch/README.md @@ -276,11 +276,11 @@ BERT的全称是Bidirectional Encoder Representation from Transformers,即双 | 1p-NPU | - | 171.687 | 3 | O2 | 2.1 | | 8p-NPU | 0.59 | 1357.106 | 3 | O2 | 2.1 | -说明:上表为历史数据,仅供参考。2024年12月31日更新的性能数据如下: +说明:上表为历史数据,仅供参考。2025年5月10日更新的性能数据如下: | NAME | 精度类型 | FPS | | :------: |:-------:|:------:| | 8p-竞品 | FP16 | 898 | -| 8p-Atlas 900 A2 PoDc | FP16 | 1634.74 | +| 8p-Atlas 900 A2 PoDc | FP16 | 1573.49 | # 版本说明 diff --git a/PyTorch/built-in/nlp/Fairseq_Transformer_wmt18_for_PyTorch/README.md b/PyTorch/built-in/nlp/Fairseq_Transformer_wmt18_for_PyTorch/README.md index e55a929c8c..dbfc6d607b 100644 --- a/PyTorch/built-in/nlp/Fairseq_Transformer_wmt18_for_PyTorch/README.md +++ b/PyTorch/built-in/nlp/Fairseq_Transformer_wmt18_for_PyTorch/README.md @@ -212,11 +212,11 @@ Fairseq Transformer wmt18模型是Fairseq套件中基于Transformer结构的翻 | 8p-NPU | fp32 | 41.04 | 223k | 20 | - | 1.11 | | 8p-NPU | fp32 | 41.04 | 221k | 20 | - | 2.1 | -说明:上表为历史数据,仅供参考。2024年12月31日更新的性能数据如下: +说明:上表为历史数据,仅供参考。2025年5月10日更新的性能数据如下: | NAME | 精度类型 | FPS | | :------ |:-------:|:------:| | 8p-竞品 | FP16 | 450000 | -| 8p-Atlas 900 A2 PoDc | FP16 | 320000 | +| 8p-Atlas 900 A2 PoDc | FP16 | 301266 | > **说明:** >由于该模型默认开启二进制,所以在性能测试时,需要安装二进制包,安装方式参考《[Pytorch框架训练环境准备](https://www.hiascend.com/document/detail/zh/ModelZoo/pytorchframework/ptes)》。 diff --git a/PyTorch/built-in/rl/MAPPO_for_PyTorch/README.md b/PyTorch/built-in/rl/MAPPO_for_PyTorch/README.md index a394db5087..949b87551f 100644 --- a/PyTorch/built-in/rl/MAPPO_for_PyTorch/README.md +++ b/PyTorch/built-in/rl/MAPPO_for_PyTorch/README.md @@ -137,11 +137,11 @@ | 1p-竞品V | 1789 | 2000000 | -15.9 | | 1p-NPU | 885 | 2000000 | -15.9 | -说明:上表为历史数据,仅供参考。2024年12月31日更新的性能数据如下: +说明:上表为历史数据,仅供参考。2025年5月10日更新的性能数据如下: | NAME | 精度类型 | FPS | | :------ |:-------:|:------:| | 1p-竞品 | FP16 | 1789 | -| 1p-Atlas 900 A2 PoDc | FP16 | 2005.73 | +| 1p-Atlas 900 A2 PoDc | FP16 | 1791.33 | # 公网地址说明 代码涉及公网地址参考 public_address_statement.md diff --git a/PyTorch/built-in/rl/PPO_for_Pytorch/README.md b/PyTorch/built-in/rl/PPO_for_Pytorch/README.md index afbfc980de..2374668751 100644 --- a/PyTorch/built-in/rl/PPO_for_Pytorch/README.md +++ b/PyTorch/built-in/rl/PPO_for_Pytorch/README.md @@ -136,11 +136,11 @@ | 1p-竞品V | 585.37 | 3000000 | 197.75 | | 1p-NPU-Atlas 800T A2 | 284.02 | 3000000 | 256.06 | -说明:上表为历史数据,仅供参考。2024年12月31日更新的性能数据如下: +说明:上表为历史数据,仅供参考。2025年5月10日更新的性能数据如下: | NAME | 精度类型 | FPS | | :------ |:-------:|:------:| | 1p-竞品 | FP16 | 585.37 | -| 1p-Atlas 900 A2 PoDc | FP16 | 489.8 | +| 1p-Atlas 900 A2 PoDc | FP16 | 413.79 | # 公网地址说明 无。 diff --git a/PyTorch/contrib/audio/wav2vec2.0/README.md b/PyTorch/contrib/audio/wav2vec2.0/README.md index 70b7ae45e0..e9a760d417 100644 --- a/PyTorch/contrib/audio/wav2vec2.0/README.md +++ b/PyTorch/contrib/audio/wav2vec2.0/README.md @@ -175,11 +175,11 @@ Wav2vec2.0是Meta在2020年发表的无监督语音预训练模型。它的核 | 1p-NPU | - | 7373.43 | - | - | 2.1 | | 8p-NPU | 5.546 | 66346.9 | - | - | 2.1 | -说明:上表为历史数据,仅供参考。2024年12月31日更新的性能数据如下: +说明:上表为历史数据,仅供参考。2025年5月10日更新的性能数据如下: | NAME | 精度类型 | FPS | | :------ |:-------:|:------:| | 8p-竞品 | FP32 | 44493.3 | -| 8p-Atlas 900 A2 PoDc | FP32 | 78111.7 | +| 8p-Atlas 900 A2 PoDc | FP32 | 77204.6 | # 版本说明 diff --git a/PyTorch/contrib/cv/classification/HRNet_ID1780_for_PyTorch/README.md b/PyTorch/contrib/cv/classification/HRNet_ID1780_for_PyTorch/README.md index 77711ae95d..bf0b579762 100644 --- a/PyTorch/contrib/cv/classification/HRNet_ID1780_for_PyTorch/README.md +++ b/PyTorch/contrib/cv/classification/HRNet_ID1780_for_PyTorch/README.md @@ -188,11 +188,11 @@ HRNet,是一个用于图像分类的高分辨网络。通过并行连接高分 | 1p-NPU | - | 773.9 | 1 | O2 | 2.1 | | 8p-NPU | 76.65 | 4856.1 | 100 | O2 | 2.1 | -说明:上表为历史数据,仅供参考。2024年12月31日更新的性能数据如下: +说明:上表为历史数据,仅供参考。2025年5月10日更新的性能数据如下: | NAME | 精度类型 | FPS | | :------ |:-------:|:------:| | 8p-竞品 | HF32 | 4856.1 | -| 8p-Atlas 900 A2 PoDc | HF32 | 3030.5 | +| 8p-Atlas 900 A2 PoDc | HF32 | 2962.7 | # 版本说明 diff --git a/PyTorch/contrib/cv/classification/InceptionV3_ID1596_for_PyTorch/README.md b/PyTorch/contrib/cv/classification/InceptionV3_ID1596_for_PyTorch/README.md index 8859be37f9..8f0395a6d8 100644 --- a/PyTorch/contrib/cv/classification/InceptionV3_ID1596_for_PyTorch/README.md +++ b/PyTorch/contrib/cv/classification/InceptionV3_ID1596_for_PyTorch/README.md @@ -193,11 +193,11 @@ GoogLeNet对网络中的传统卷积层进行了修改,提出了被称为Incep | 1p-NPU | - | 1166.723 | 1 | O2 | 2.1 | | 8p-NPU | 78.12 | 8812.33 | 240 | O2 | 2.1 | -说明:上表为历史数据,仅供参考。2024年12月31日更新的性能数据如下: +说明:上表为历史数据,仅供参考。2025年5月10日更新的性能数据如下: | NAME | 精度类型 | FPS | | :------ |:-------:|:------:| | 8p-竞品 | FP16 | 4995.12 | -| 8p-Atlas 900 A2 PoDc | FP16 | 9641.58 | +| 8p-Atlas 900 A2 PoDc | FP16 | 8666.48 | # 版本说明 diff --git a/PyTorch/contrib/cv/classification/MAE_for_PyTorch/README.md b/PyTorch/contrib/cv/classification/MAE_for_PyTorch/README.md index 88b8ae41a6..1ad23f5e59 100644 --- a/PyTorch/contrib/cv/classification/MAE_for_PyTorch/README.md +++ b/PyTorch/contrib/cv/classification/MAE_for_PyTorch/README.md @@ -235,11 +235,11 @@ MAE的设计虽然简单,但已被证明是一个强大的、可扩展的视 说明1:MAE-Large配置在910上进行训练 -说明2:上表为历史数据,仅供参考。2024年12月31日更新的性能数据如下: +说明2:上表为历史数据,仅供参考。2025年5月10日更新的性能数据如下: | NAME | 精度类型 | FPS | | :------ |:-------:|:------:| | 8p-竞品 | FP16 | 3538.01 | -| 8p-Atlas 900 A2 PoDc | FP16 | 3769.55 | +| 8p-Atlas 900 A2 PoDc | FP16 | 3874.48 | # 版本说明 diff --git a/PyTorch/contrib/cv/pose_estimation/Lightweight_OpenPose/README.md b/PyTorch/contrib/cv/pose_estimation/Lightweight_OpenPose/README.md index 8b8efca2ee..733e4a6839 100644 --- a/PyTorch/contrib/cv/pose_estimation/Lightweight_OpenPose/README.md +++ b/PyTorch/contrib/cv/pose_estimation/Lightweight_OpenPose/README.md @@ -263,11 +263,11 @@ step-3阶段结果 **说明1:** 以上精度结果为执行train_full_8p.sh脚本所得的验证集的结果。 -说明2:上表为历史数据,仅供参考。2024年12月31日更新的性能数据如下: +说明:上表为历史数据,仅供参考。2025年5月10日更新的性能数据如下: | NAME | 精度类型 | FPS | | :------ |:-------:|:------:| | 8p-竞品 | FP16 | 3005.633 | -| 8p-Atlas 900 A2 PoDc | FP16 | 4336.99 | +| 8p-Atlas 900 A2 PoDc | FP16 | 3897.37 | # 版本说明 diff --git a/PyTorch/contrib/cv/semantic_segmentation/DeeplabV3_for_Pytorch/README.md b/PyTorch/contrib/cv/semantic_segmentation/DeeplabV3_for_Pytorch/README.md index 3a30d8d284..7b28d6bc14 100644 --- a/PyTorch/contrib/cv/semantic_segmentation/DeeplabV3_for_Pytorch/README.md +++ b/PyTorch/contrib/cv/semantic_segmentation/DeeplabV3_for_Pytorch/README.md @@ -209,11 +209,11 @@ DeepLabV3是一个经典的语义分割网络,采用空洞卷积来代替池 | 1p-NPU | - | - | 14.806 | 1000 | O2 | 1.11 | | 8p-NPU | 96.13 | 78.98 | 115.593 | 7000 | O2 | 1.11 | -说明:上表为历史数据,仅供参考。2024年12月31日更新的性能数据如下: +说明:上表为历史数据,仅供参考。2025年5月10日更新的性能数据如下: | NAME | 精度类型 | FPS | | :------ |:-------:|:------:| | 8p-竞品 | FP16 | 115.593 | -| 8p-Atlas 900 A2 PoDc | FP16 | 147 | +| 8p-Atlas 900 A2 PoDc | FP16 | 138.92 | # 版本说明 diff --git a/PyTorch/contrib/nlp/BERT-NER-Pytorch/README.md b/PyTorch/contrib/nlp/BERT-NER-Pytorch/README.md index a4a8a7d8f4..8dd48a2c65 100644 --- a/PyTorch/contrib/nlp/BERT-NER-Pytorch/README.md +++ b/PyTorch/contrib/nlp/BERT-NER-Pytorch/README.md @@ -127,11 +127,11 @@ BERT-NER-Pytorch | --- | --- |---------| --- | --- | | 8p-NPU | 79.16 | 1163.21 | 1129.4 | 4 | -说明:上表为历史数据,仅供参考。2024年12月31日更新的性能数据如下: +说明:上表为历史数据,仅供参考。2025年5月10日更新的性能数据如下: | NAME | 精度类型 | FPS | | :------ |:-------:|:------:| | 8p-竞品 | FP32 | 1942.15 | -| 8p-Atlas 900 A2 PoDc | FP32 | 1662.19 | +| 8p-Atlas 900 A2 PoDc | FP32 | 1407.21 | ## 版本说明 ### 变更 diff --git a/PyTorch/contrib/nlp/albert_ID0335_for_PyTorch/README.md b/PyTorch/contrib/nlp/albert_ID0335_for_PyTorch/README.md index e1249c0ad2..e530265ece 100644 --- a/PyTorch/contrib/nlp/albert_ID0335_for_PyTorch/README.md +++ b/PyTorch/contrib/nlp/albert_ID0335_for_PyTorch/README.md @@ -185,11 +185,11 @@ Albert是自然语言处理模型,基于Bert模型修改得到。相比于Bert | 1p-NPU | 0.932 | 1025.36 | 2 | O2 | 2.1 | | 8p-NPU | 0.927 | 6394.05 | 7 | O2 | 2.1 | -说明:上表为历史数据,仅供参考。2024年12月31日更新的性能数据如下: +说明:上表为历史数据,仅供参考。2025年5月10日更新的性能数据如下: | NAME | 精度类型 | FPS | | :------: |:-------:|:------:| | 8p-竞品 | FP16 | 6394.05 | -| 8p-Atlas 900 A2 PoDc | FP16 | 8970.89 | +| 8p-Atlas 900 A2 PoDc | FP16 | 9186.47 | # 版本说明 diff --git a/PyTorch/contrib/nlp/roberta_for_PyTorch/README.md b/PyTorch/contrib/nlp/roberta_for_PyTorch/README.md index 0bc3684031..bfd687a9f4 100644 --- a/PyTorch/contrib/nlp/roberta_for_PyTorch/README.md +++ b/PyTorch/contrib/nlp/roberta_for_PyTorch/README.md @@ -191,11 +191,11 @@ RoBERTa 在模型规模、算力和数据上,都比 BERT 有一定的提升。 | 1p-NPU | 0.938 | 879.05 | 1 | O2 | 2.1 | | 8p-NPU | 0.969 | 7078.64 | 10 | O2 | 2.1 | -说明:上表为历史数据,仅供参考。2024年12月31日更新的性能数据如下: +说明:上表为历史数据,仅供参考。2025年5月10日更新的性能数据如下: | NAME | 精度类型 | FPS | | :------ |:-------:|:------:| | 8p-竞品 | FP16 | 7309.36 | -| 8p-Atlas 900 A2 PoDc | FP16 | 8084.65 | +| 8p-Atlas 900 A2 PoDc | FP16 | 7214.09 | # 版本说明 diff --git a/PyTorch/dev/perf/ShuffleNetV2_iflytek_for_Pytorch/README.md b/PyTorch/dev/perf/ShuffleNetV2_iflytek_for_Pytorch/README.md index 583e969344..e9e1620395 100644 --- a/PyTorch/dev/perf/ShuffleNetV2_iflytek_for_Pytorch/README.md +++ b/PyTorch/dev/perf/ShuffleNetV2_iflytek_for_Pytorch/README.md @@ -191,11 +191,11 @@ ShuffleNetV2是一个改进ShuffleNetV1的轻量级的网络,为了解决在 | 1p-NPU | - | 2081.3 | 90 | FP32 | 2.1 | | 8p-NPU | 62.714 | 10502.56 | 90 | FP32 | 2.1 | -说明:上表为历史数据,仅供参考。2024年12月31日更新的性能数据如下: +说明:上表为历史数据,仅供参考。2025年5月10日更新的性能数据如下: | NAME | 精度类型 | FPS | | :------ |:-------:|:------:| | 8p-竞品 | FP32 | 10502.56 | -| 8p-Atlas 900 A2 PoDc | FP32 | 15170.37 | +| 8p-Atlas 900 A2 PoDc | FP32 | 14524.82 | # 版本说明 -- Gitee