Add kunlunxin mask-rcnn (#276)

* Add kunlunxin mask-rcnn * Refine mask-rcnn --------- Co-authored-by: root <[email protected]>
FlagOpen · Oct 9, 2023 · fbba360 · fbba360
1 parent 903d60b
commit fbba360
Show file tree

Hide file tree

Showing 7 changed files with 82 additions and 0 deletions.
diff --git a/training/kunlunxin/mask_rcnn-pytorch/README.md b/training/kunlunxin/mask_rcnn-pytorch/README.md
@@ -0,0 +1,48 @@
+### 模型Checkpoint下载
+[模型Checkpoint下载](../../benchmarks/mask_rcnn/README.md#模型checkpoint)
+### 测试数据集下载
+[测试数据集下载](../../benchmarks/mask_rcnn/README.md#数据集下载地址)
+
+### 昆仑芯XPU配置与运行信息参考
+#### 环境配置
+- ##### 硬件环境
+  - 机器型号: 昆仑芯AI加速器组R480-X8
+  - 加速卡型号: 昆仑芯AI加速卡R300
+  - 多机网络类型、带宽: InfiniBand，200Gb/s
+
+- ##### 软件环境
+  - OS版本：Ubuntu 20.04
+  - OS kernel版本: 5.4.0-26-generic
+  - 加速卡驱动版本：4.0.25
+  - Docker镜像和版本：pytorch1.12.1-cpu-ubuntu20.04:v0.01
+  - 训练框架版本：xmlir
+  - 训练编译器版本：xacc
+  - 依赖软件版本：pytorch-1.12.1+cpu
+
+
+### 运行情况
+* 通用指标
+
+| 指标名称       | 指标值                  | 特殊说明                                    |
+| -------------- | ----------------------- | ------------------------------------------- |
+| 任务类别       | 图像目标检测及语义分割            |                                             |
+| 模型           | mask_rcnn               |                                             |
+| 数据集         | coco2017                |                                             |
+| 数据精度       | precision,见“性能指标”  | 可选fp32/amp/fp16                           |
+| 超参修改       | fix_hp,见“性能指标”     | 跑满硬件设备评测吞吐量所需特殊超参          |
+| 硬件设备简称   | R300            |                                             |
+| 硬件存储使用   | mem,见“性能指标”        | 通常称为“显存”,单位为GiB                    |
+| 端到端时间     | e2e_time,见“性能指标”   | 总时间+Perf初始化等时间                     |
+| 总吞吐量       | p_whole,见“性能指标”    | 实际训练图片数除以总时间(performance_whole) |
+| 训练吞吐量     | p_train,见“性能指标”    | 不包含每个epoch末尾的评估部分耗时           |
+| **计算吞吐量** | **p_core,见“性能指标”** | 不包含数据IO部分的耗时(p3>p2>p1)            |
+| 训练结果       | map,见“性能指标”        | 单位为平均目标检测正确率                    |
+| 额外修改项     | 无                      |                                             |
+
+* 性能指标
+
+| 配置                | precision | fix_hp        | e2e_time | p_whole | p_train | p_core | map_bbox && map_segm | mem       |
+| ------------------- | --------- | ------------- | -------- | ------- | ------- | ------ | -------------------- | --------- |
+| R300单机单卡（1x1） | fp32      |   |          |         |         |         |             |            |
+| R300单机8卡（1x8）  | fp32      | bs=8,lr=0.16   |           |        |        |           |  38.4 && 34.6   |   11.3/32.0  |
+| R300两机8卡（2x8）  | fp32      |    |           |         |         |         |             |            |
diff --git a/training/kunlunxin/mask_rcnn-pytorch/config/config_R300x1x1.py b/training/kunlunxin/mask_rcnn-pytorch/config/config_R300x1x1.py
@@ -0,0 +1,7 @@
+vendor: str = "kunlunxin"
+
+distributed = False
+
+train_batch_size = 8
+eval_batch_size = 8
+lr = 0.16
diff --git a/training/kunlunxin/mask_rcnn-pytorch/config/config_R300x1x8.py b/training/kunlunxin/mask_rcnn-pytorch/config/config_R300x1x8.py
@@ -0,0 +1,7 @@
+vendor: str = "kunlunxin"
+
+dist_backend = "xccl"
+
+train_batch_size = 8
+eval_batch_size = 8
+lr = 0.16
diff --git a/training/kunlunxin/mask_rcnn-pytorch/config/config_R300x2x8.py b/training/kunlunxin/mask_rcnn-pytorch/config/config_R300x2x8.py
@@ -0,0 +1,7 @@
+vendor: str = "kunlunxin"
+
+dist_backend = "xccl"
+
+train_batch_size = 8
+eval_batch_size = 8
+lr = 0.16
diff --git a/training/kunlunxin/mask_rcnn-pytorch/config/environment_variables.sh b/training/kunlunxin/mask_rcnn-pytorch/config/environment_variables.sh
@@ -0,0 +1,9 @@
+# =================================================
+# Export variables
+# =================================================
+
+
+export XACC_ENABLE=1
+export BKCL_PCIE_RING=1
+export XMLIR_D_XPU_L3_SIZE=66060288
+export XDNN_CONV_GEMM_DTYPE="int16"
diff --git a/training/kunlunxin/mask_rcnn-pytorch/config/requirements.txt b/training/kunlunxin/mask_rcnn-pytorch/config/requirements.txt
@@ -0,0 +1,4 @@
+https://download.pytorch.org/whl/cpu/torchvision-0.13.1%2Bcpu-cp38-cp38-linux_x86_64.whl
+pycocotools==2.0.5
+psutil==5.9.5
+numpy==1.23.5
diff --git a/training/kunlunxin/mask_rcnn-pytorch/extern/.gitkeep b/training/kunlunxin/mask_rcnn-pytorch/extern/.gitkeep