whuluojiateam/luojianet

LuoJiaNet介绍

遥感专用机器学习框架LuoJiaNET,由武汉大学LuoJiaNET框架团队与华为MindSpore框架研究小组联合打造而成， 是遥感领域首个国产化自主可控的遥感专用机器学习框架，针对遥感数据像幅尺寸大、数据通道多、尺度变化大等特性， 具备内存可扩展、尺度通道灵活创建、数据通道自主优选、框架与数据协同处理的特点。可兼容已有深度学习框架， 并提供用户友好的、可拖拽的交互式网络结构搭建界面的方法。能屏蔽不同硬件设备间差异，同时管理多样化的遥感影像样本库LuoJiaSET， 实现遥多源感影像样本的高效存储管理。

LuoJiaNET同时与国产人工智能硬件NPU深度融合，可同时支持CPU、GPU、NPU等硬件资源，使智能计算软硬件充分协同，形成融合探测机理与地学知识的统一计算图表达、编译优化、图算融合、 自动混合并行的新一代遥感智能解译框架，可进行遥感样本自动提纯与增广，充分融合探测机理与地学知识。

1. 请参照LuoJiaNet安装手册 安装whl包或源码编译安装

2. 执行以下命令，验证安装结果。

   import numpy as np
   import luojianet_ms.context as context
   import luojianet_ms.nn as nn
   from luojianet_ms import Tensor
   from luojianet_ms.ops import operations as P
   
   context.set_context(mode=context.GRAPH_MODE, device_target="CPU")
   
   class Mul(nn.Module):
       def __init__(self):
           super(Mul, self).__init__()
           self.mul = P.Mul()
   
       def forward(self, x, y):
           return self.mul(x, y)
   
   x = Tensor(np.array([1.0, 2.0, 3.0]).astype(np.float32))
   y = Tensor(np.array([4.0, 5.0, 6.0]).astype(np.float32))
   
   mul = Mul()
   print(mul(x, y))

   [ 4. 10. 18.]
   

3. LuoJiaNET安装手册中，相应的源码见tutorial

版本说明

版本说明请参阅RELEASE。

许可证

Apache License 2.0

luojianet's Issues

如何在Ascend710训练服务器安装LuojiaNET

LuojiaNET团队您好

查看官网安装信息，发现仅提供x86 cpu以及gpu环境whl安装文件。请问如何在华为Ascend训练服务器安装LuojiaNET。

感激回答。

Mnist dataset test error

Compling with cuda 11.1 and cudnn 8 on unbuntu 18.04, run

example/quick_start.py

got error of

RuntimeError: Unexpected error. Invalid mnist file, the image number of /home/work/luojia/luojianet/datasets/MNIST_Data/train/train-images-idx3-ubyte should be 2051, but got 1010792557
Line of code : 119
File         : /home/work/luojia/luojianet/luojianet_ms/ccsrc/minddata/dataset/engine/datasetops/source/mnist_op.cc

It looks like dataset loading problem.

I had this problem compiling GDAL when using Windows

[BUG]训练model_zoo/rs_semantic_segmentation下面的网络做二分类任务时，总遇到labe的维度不一致

训练model_zoo/rs_semantic_segmentation下面的网络做二分类任务时，总遇到labe的维度不一致，请问一下怎么解决

[BUG]使用model_zoo里训练resnet时，报错RuntimeError: Unexpected error. Failed to open file: ./WHU-RS19/Parking/parking_28.jpg

============== Starting Training ==============
[WARNING] ME(19208:139772197278208,MainProcess):2022-10-27-18:32:37.851.63 [luojianet_ms/train/model.py:573] The CPU cannot support dataset sink mode currently.So the training process will be performed with dataset not sink.
[WARNING] DEBUG(19208,7f1f402b1200,python):2022-10-27-18:32:37.123.286 [luojianet_ms/ccsrc/debug/debugger/debugger.cc:96] Debugger] Not enabling debugger. Debugger does not support CPU.
[ERROR] MD(19208,7f1e8fbba700,python):2022-10-27-18:32:37.156.100 [luojianet_ms/ccsrc/minddata/dataset/util/task_manager.cc:219] InterruptMaster] Task is terminated with err msg(more detail in info level log):Unexpected error. Failed to open file: ./WHU-RS19/Parking/parking_28.jpg
Line of code : 277
File : /home/vhr/tll/download/luojianet/luojianet_ms/ccsrc/minddata/dataset/core/tensor.cc

[WARNING] ME(19709:140164741772224,MainProcess):2022-10-27-18:32:38.345.906 [luojianet_ms/common/_decorator.py:34] 'DropoutDoMask' is deprecated from version 1.5 and will be removed in a future version, use 'ops.Dropout' instead.
The new bprop mindir files has been generated in the path "/home/riemann/anaconda3/envs/luojianet/lib/python3.9/site-packages/luojianet_ms/ops/_grad/../bprop_mindir", copy the *.mindir to your luojianet_ms path or PYTHONPATH if necessary.
epoch: 1 step: 1, loss is 2.9637277126312256
Traceback (most recent call last):
File "/home/y00572966/luojianet_test/luojianet/model_zoo/rs_scene_classification/ResNet/train.py", line 77, in
model.train(config.epoch_size, dataset, callbacks=[time_cb,ckpoint_cb,LossMonitor()])
File "/home/riemann/anaconda3/envs/luojianet/lib/python3.9/site-packages/luojianet_ms/train/model.py", line 808, in train
self._train(epoch,
File "/home/riemann/anaconda3/envs/luojianet/lib/python3.9/site-packages/luojianet_ms/train/model.py", line 112, in wrapper
func(self, *args, **kwargs)
File "/home/riemann/anaconda3/envs/luojianet/lib/python3.9/site-packages/luojianet_ms/train/model.py", line 575, in _train
self._train_process(epoch, train_dataset, list_callback, cb_params)
File "/home/riemann/anaconda3/envs/luojianet/lib/python3.9/site-packages/luojianet_ms/train/model.py", line 693, in _train_process
for next_element in dataset_helper:
File "/home/riemann/anaconda3/envs/luojianet/lib/python3.9/site-packages/luojianet_ms/train/dataset_helper.py", line 523, in next
data = self.iter.next()
File "/home/riemann/anaconda3/envs/luojianet/lib/python3.9/site-packages/luojianet_ms/dataset/engine/iterators.py", line 144, in next
data = self._get_next()
File "/home/riemann/anaconda3/envs/luojianet/lib/python3.9/site-packages/luojianet_ms/dataset/engine/iterators.py", line 222, in _get_next
return [self._transform_tensor(t) for t in self._iterator.GetNextAsList()]
RuntimeError: Unexpected error. Failed to open file: ./WHU-RS19/Parking/parking_28.jpg
Line of code : 277
File : /home/vhr/tll/download/luojianet/luojianet_ms/ccsrc/minddata/dataset/core/tensor.cc

luojianet针对大幅面遥感影像解译的模块，具体怎么使用的

luojianet团队您好：
有幸接触到luojianet这样一个遥感专用机器学习框架，我对其中的大幅面遥感影像解译比较感兴趣。查阅了相关资料，还是不太清楚在使用时，具体应该怎么做。具体问题是：
在模型训练的时候，是将整幅影像扔进去训练吗，不需要裁切成小图了是吗？
扔进去之后，相较于常规的深度学习框架中的模型训练，会有哪些不一样的处理呢？
希望您能帮我解答疑惑

编译GDAL，执行中出现recipe for target 'autotest/cpp/libvsipreload.so' failed问题，如何解决

系统环境介绍：
ubuntu18.04, cuda11.1, cudnn8.8.0，python3.7.6, gcc7.5.0
按luojianet官网安装流程进行编译安装。
在安装过程中，进行到此流程。
编译GDAL， 执行bash build_gdal.sh出现如下错误：
[ 93%] Built target gnmanalyse
[ 93%] Building CXX object apps/CMakeFiles/gnmmanage.dir/gnmmanage.cpp.o
[ 93%] Linking CXX executable gnmmanage
[ 93%] Built target gnmmanage
[ 93%] Building CXX object apps/CMakeFiles/test_ogrsf.dir/test_ogrsf.cpp.o
[ 93%] Linking CXX executable test_ogrsf
[ 93%] Built target test_ogrsf
[ 93%] Building CXX object autotest/cpp/CMakeFiles/vsipreload.dir///port/vsipreload.cpp.o
[ 93%] Linking CXX shared module libvsipreload.so
CMakeFiles/vsipreload.dir///port/vsipreload.cpp.o: In function myinit()': vsipreload.cpp:(.text+0x92): undefined reference to dlsym'
vsipreload.cpp:(.text+0xac): undefined reference to dlsym' vsipreload.cpp:(.text+0xc6): undefined reference to dlsym'
vsipreload.cpp:(.text+0xe0): undefined reference to dlsym' vsipreload.cpp:(.text+0xfa): undefined reference to dlsym'
CMakeFiles/vsipreload.dir///port/vsipreload.cpp.o:vsipreload.cpp:(.text+0x114): more undefined references to `dlsym' follow
collect2: error: ld returned 1 exit status
autotest/cpp/CMakeFiles/vsipreload.dir/build.make:97: recipe for target 'autotest/cpp/libvsipreload.so' failed
make[2]: *** [autotest/cpp/libvsipreload.so] Error 1
CMakeFiles/Makefile2:9406: recipe for target 'autotest/cpp/CMakeFiles/vsipreload.dir/all' failed
make[1]: *** [autotest/cpp/CMakeFiles/vsipreload.dir/all] Error 2
Makefile:145: recipe for target 'all' failed
make: *** [all] Error 2

不知道什么原因造成的，希望luojianet团队人员帮忙解答一下，谢谢！！！

源码编译遇到的错误

按照官方教程一步步来的，1.0.6和master版本都试过了，都是同样的错误，到下面这一步报错，貌似跟openmpi有关

......
......
checking for netinet/tcp.h... yes
checking for struct sockaddr_in... yes
checking if --with-cuda is set... not set (--with-cuda=)
./configure: line 13028: syntax error near unexpected token )' ./configure: line 13028: )'
CMake Error at cmake/utils.cmake:179 (message):
error! when ./configure;CXXFLAGS=-D_FORTIFY_SOURCE=2
-O2;--prefix=/home/rs/Codes/luojianet/build/luojianet_ms/.mslib/ompi_4.0.3_653b828d1fcd585d80aa35947b9dfeae
in /home/rs/Codes/luojianet/build/luojianet_ms/_deps/ompi-src
Call Stack (most recent call first):
cmake/utils.cmake:393 (__exec_cmd)
cmake/external_libs/ompi.cmake:10 (luojianet_ms_add_pkg)
cmake/mind_expression.cmake:43 (include)
CMakeLists.txt:56 (include)

-- Configuring incomplete, errors occurred!
See also "/home/rs/Codes/luojianet/build/luojianet_ms/CMakeFiles/CMakeOutput.log".
See also "/home/rs/Codes/luojianet/build/luojianet_ms/CMakeFiles/CMakeError.log".

[BUG]model zoo中deeplabv3训练中出现RuntimeError: luojianet_ms/ccsrc/frontend/optimizer/ad/kprim.cc:199 ImportBpropFromMindIR] The bprop mindir files are not up to da te.

deeplabv3模型训练时出现以下错误：
[WARNING] ME(2199:140703017723712,MainProcess):2022-10-28-16:29:18.609.789 [luojianet_ms/common/_decorator.py:38] 'TensorAdd' is d eprecated from version 1.1 and will be removed in a future version, use 'Add' instead.
Total Epoch:200 Training num:10000 Validation num:1000
INFO:log:Total Epoch:200 Training num:10000 Validation num:1000
[ERROR] OPTIMIZER(2199,7ff7f963e740,python3):2022-10-28-16:29:25.897.388 [luojianet_ms/ccsrc/frontend/optimizer/ad/kprim.cc:172] C heckBpropHash] The bprop mindir files are not up to date. Please run the /usr/local/python3.7.5/lib/python3.7/site-packages/luojia net_ms/ops/_grad/../bprop_mindir/generate_mindir.py to generate new mindir files.
bprop_fg hash: 3d4ca3af3054d32fe54a557e457674558c4179705eccb4c3dae775993ba1a76a
bprop hash list:

[ERROR] OPTIMIZER(2199,7ff7f963e740,python3):2022-10-28-16:29:25.897.420 [luojianet_ms/ccsrc/frontend/optimizer/ad/kprim.cc:177] C heckBpropHash] 5dbd9e9d72c7b3227bfe7cc41cc4311526259c3297cb24ab2d4d7aa122c901e6
[ERROR] OPTIMIZER(2199,7ff7f963e740,python3):2022-10-28-16:29:25.897.430 [luojianet_ms/ccsrc/frontend/optimizer/ad/kprim.cc:177] C heckBpropHash] d15b0d6de0f996dcacce657aaec10a3a526e8567314332c8628d9cbc7bc26a03
[ERROR] OPTIMIZER(2199,7ff7f963e740,python3):2022-10-28-16:29:25.897.438 [luojianet_ms/ccsrc/frontend/optimizer/ad/kprim.cc:177] C heckBpropHash] c252daaf204cd7d19bad3054c4734cd874f00b9a0ad6d8b3b01ffe56bf1f0b2f
[ERROR] OPTIMIZER(2199,7ff7f963e740,python3):2022-10-28-16:29:25.897.448 [luojianet_ms/ccsrc/frontend/optimizer/ad/kprim.cc:177] C heckBpropHash] 15cf11818e324bb31ab387f78edcce13902e54eb561b8406f4aac743e626cc2a
[ERROR] OPTIMIZER(2199,7ff7f963e740,python3):2022-10-28-16:29:25.897.455 [luojianet_ms/ccsrc/frontend/optimizer/ad/kprim.cc:177] C heckBpropHash] da9b8c2c77b9335db3c701e0a27fc1b3514c7abf3614b2b477cf94d2420d770e
[ERROR] OPTIMIZER(2199,7ff7f963e740,python3):2022-10-28-16:29:25.897.461 [luojianet_ms/ccsrc/frontend/optimizer/ad/kprim.cc:177] C heckBpropHash] dcbcbe9c73bbeb292f97cefef13d1e66d4df58dc4d0f4d61b3b3c0c48e61f014
[ERROR] OPTIMIZER(2199,7ff7f963e740,python3):2022-10-28-16:29:25.897.466 [luojianet_ms/ccsrc/frontend/optimizer/ad/kprim.cc:177] C heckBpropHash] 300e3a12184504bf922c740d3c92af822c2517aef91dd32c009b58266169f93d
[ERROR] OPTIMIZER(2199,7ff7f963e740,python3):2022-10-28-16:29:25.897.473 [luojianet_ms/ccsrc/frontend/optimizer/ad/kprim.cc:177] C heckBpropHash] 8a2f1d24f72f3e2821cdd873682fda4b8322905ab83ebe12885e2c9aec2957d0
[ERROR] OPTIMIZER(2199,7ff7f963e740,python3):2022-10-28-16:29:25.897.478 [luojianet_ms/ccsrc/frontend/optimizer/ad/kprim.cc:177] C heckBpropHash] 9c4e884cbc35fd59fd028a19bc7f9c9510d0b05d90ad3de2939e2498a66a6ade
[ERROR] OPTIMIZER(2199,7ff7f963e740,python3):2022-10-28-16:29:25.897.484 [luojianet_ms/ccsrc/frontend/optimizer/ad/kprim.cc:177] C heckBpropHash] fd36050ed51ca754e75ba8ab0e59cf7aefc9aaf4425e607aa1d6115a69d64933
[ERROR] OPTIMIZER(2199,7ff7f963e740,python3):2022-10-28-16:29:25.897.488 [luojianet_ms/ccsrc/frontend/optimizer/ad/kprim.cc:177] C heckBpropHash] 55e08f68aa4e972c4ebb7fc191119b682d6347830d383020968203382a8b59cc
[ERROR] OPTIMIZER(2199,7ff7f963e740,python3):2022-10-28-16:29:25.897.493 [luojianet_ms/ccsrc/frontend/optimizer/ad/kprim.cc:177] C heckBpropHash] ec3beb004dc8c56197a40e8b91c092539a103b8e7e7be62edaf0e83671e1966e
[ERROR] OPTIMIZER(2199,7ff7f963e740,python3):2022-10-28-16:29:25.897.499 [luojianet_ms/ccsrc/frontend/optimizer/ad/kprim.cc:177] C heckBpropHash] 2930c5739d67414f42123aec886e845ed9e11e82b7a792124f3a04f57e17b5c4
[CRITICAL] OPTIMIZER(2199,7ff7f963e740,python3):2022-10-28-16:29:25.897.507 [luojianet_ms/ccsrc/frontend/optimizer/ad/kprim.cc:199 ] ImportBpropFromMindIR] The bprop mindir files are not up to date.
[WARNING] VM(2199,7ff768fb9700,python3):2022-10-28-16:29:25.897.759 [luojianet_ms/ccsrc/runtime/pynative/op_task.h:106] Run] Op bu ild failed, no need to launch.
Traceback (most recent call last):
File "train.py", line 67, in
train_net()
File "train.py", line 63, in train_net
train(param=param, model=model, train_dataset=train_dataset, valid_dataset=val_dataset)
File "/code/utils/deeplearning_dp.py", line 121, in train
train_net_step(data, label)
File "/usr/local/python3.7.5/lib/python3.7/site-packages/luojianet_ms/nn/cell.py", line 613, in call
raise err
File "/usr/local/python3.7.5/lib/python3.7/site-packages/luojianet_ms/nn/cell.py", line 610, in call
output = self._run_construct(cast_inputs, kwargs)
File "/usr/local/python3.7.5/lib/python3.7/site-packages/luojianet_ms/nn/cell.py", line 430, in _run_construct
output = self.forward(*cast_inputs, **kwargs)
File "/code/utils/deeplearning_dp.py", line 50, in forward
loss = self.network(data, label)
File "/usr/local/python3.7.5/lib/python3.7/site-packages/luojianet_ms/nn/cell.py", line 613, in call
raise err
File "/usr/local/python3.7.5/lib/python3.7/site-packages/luojianet_ms/nn/cell.py", line 610, in call
output = self._run_construct(cast_inputs, kwargs)
File "/usr/local/python3.7.5/lib/python3.7/site-packages/luojianet_ms/nn/cell.py", line 430, in _run_construct
output = self.forward(*cast_inputs, **kwargs)
File "/code/utils/deeplearning_dp.py", line 25, in forward
out = self.backbone(data)
File "/usr/local/python3.7.5/lib/python3.7/site-packages/luojianet_ms/nn/cell.py", line 613, in call
raise err
File "/usr/local/python3.7.5/lib/python3.7/site-packages/luojianet_ms/nn/cell.py", line 610, in call
output = self._run_construct(cast_inputs, kwargs)
File "/usr/local/python3.7.5/lib/python3.7/site-packages/luojianet_ms/nn/cell.py", line 430, in _run_construct
output = self.forward(*cast_inputs, **kwargs)
File "/code/nets/init.py", line 17, in forward
x =self.model(x)
File "/usr/local/python3.7.5/lib/python3.7/site-packages/luojianet_ms/nn/cell.py", line 613, in call
raise err
File "/usr/local/python3.7.5/lib/python3.7/site-packages/luojianet_ms/nn/cell.py", line 610, in call
output = self._run_construct(cast_inputs, kwargs)
File "/usr/local/python3.7.5/lib/python3.7/site-packages/luojianet_ms/nn/cell.py", line 430, in _run_construct
output = self.forward(*cast_inputs, **kwargs)
File "/code/nets/deeplabv3.py", line 204, in forward
out = self.resnet(x)
File "/usr/local/python3.7.5/lib/python3.7/site-packages/luojianet_ms/nn/cell.py", line 613, in call
raise err
File "/usr/local/python3.7.5/lib/python3.7/site-packages/luojianet_ms/nn/cell.py", line 610, in call
output = self._run_construct(cast_inputs, kwargs)
File "/usr/local/python3.7.5/lib/python3.7/site-packages/luojianet_ms/nn/cell.py", line 430, in _run_construct
output = self.forward(*cast_inputs, **kwargs)
File "/code/nets/deeplabv3.py", line 64, in forward
out = self.relu(out)
File "/usr/local/python3.7.5/lib/python3.7/site-packages/luojianet_ms/nn/cell.py", line 613, in call
raise err
File "/usr/local/python3.7.5/lib/python3.7/site-packages/luojianet_ms/nn/cell.py", line 610, in call
output = self._run_construct(cast_inputs, kwargs)
File "/usr/local/python3.7.5/lib/python3.7/site-packages/luojianet_ms/nn/cell.py", line 430, in _run_construct
output = self.forward(*cast_inputs, **kwargs)
File "/usr/local/python3.7.5/lib/python3.7/site-packages/luojianet_ms/nn/layer/activation.py", line 299, in forward
return self.relu(x)
File "/usr/local/python3.7.5/lib/python3.7/site-packages/luojianet_ms/ops/primitive.py", line 295, in call
return _run_op(self, self.name, args)
File "/usr/local/python3.7.5/lib/python3.7/site-packages/luojianet_ms/common/api.py", line 91, in wrapper
results = fn(*arg, **kwargs)
File "/usr/local/python3.7.5/lib/python3.7/site-packages/luojianet_ms/ops/primitive.py", line 755, in _run_op
output = real_run_op(obj, op_name, args)
RuntimeError: luojianet_ms/ccsrc/frontend/optimizer/ad/kprim.cc:199 ImportBpropFromMindIR] The bprop mindir files are not up to date.
以上出现的错误，是什么原因造成的呢，希望给予帮助，非常感谢！！！

Expected Behavior

Current Behavior

Context

Steps to Reproduce

Your Environment

NPU（Ascend）服务器编译LuoJiaNET V1.0.5失败

LuoJiaTeam您好

我在华为Ascend服务器源码安装LuoJiaNET V1.0.5失败，希望LuoJiaTeam能协助分析安装失败的原因。
安装步骤：参考GPU源码编译指导。先安装环境依赖，接着安装gdal(bash build_gdal_npu.sh)，最后安装LuoJiaNET（bash build_npu.sh -e ascend -j 4）。

安装LuoJiaNET过程遇到如下问题。

问题1：采用NPU训练脚本，但编译过程会寻找CUDA
CMake Error at /usr/local/share/cmake-3.21/Modules/FindCUDA.cmake:859 (message): Specify CUDA_TOOLKIT_ROOT_DIR
规避方法： cmake/options.cmake中的option(ENABLE_GPU "Enable gpu" ON)修改为option(ENABLE_GPU "Enable gpu" OFF)

问题2：找不到头文件。经查，未在对应目录下找到指定头文件
/root/ai_framework/luojianet/graphengine/ge/ge_runtime/model_runner.cc:18:10: fatal error: framework/ge_runtime/model_runner.h: No such file or directory #include "framework/ge_runtime/model_runner.h
/root/ai_framework/luojianet/graphengine/metadef/inc/common/opskernel/ops_kernel_builder.h:25:10: fatal error: proto/task.pb.h: No such file or directory #include "proto/task.pb.h

问题3：使用的变量未定义
/root/ai_framework/luojianet/graphengine/ge/common/dump/dump_properties.cc:146:7: error: ‘mmIsDir’ was not declared in this scope
规避方法：添加 #include "mmap/mmpa_api.h"头文件。

感谢答复。

关于大幅面影像处理的问题

在官方教程中http://58.48.42.237/luojiaNet/tutorial/experts有提到大幅面处理的相关方法。
请问大幅面影像四叉树处理和大幅面影像算子分解是什么关系呢？
在代码中我只看到了QuadSearch_GID的整体流程，没有看到大幅面影像算子分解的具体代码。所以大幅面影像算子分解是四叉树处理中的一个部分吗？还是说暂时还没有大幅面影像算子分解的具体代码（教程中只有代码片段，不能直接运行）

[BUG]数据集迭代器的内存泄露问题

数据集的迭代器不能在For循环中创建，否则会带来不正常的内存回收问题，导致主存占用持续增加。
问题发现于LuojiaNet modelzoo中rs_semantic_segmentation/DeepLabv3_ISPRS_Vaihingen。代码中没有使用model.train()封装方式，而是自己编写epoch循环，使用TrainOneStepCell方式来训练网络，这时就需要手动创建数据集的迭代器来读取数据。
但如果在Epoch的For循环中每次创建迭代器，如下所示，则会有内存泄漏问题，主存会随着Epoch的增加而增长。

for epoch in range(epochs):
    for batch_samples in train_loader.create_dict_iterator():
        pass

暂时是通过在For循环外定义迭代器解决此问题，如下所示。

train_iterator = train_loader.create_dict_iterator()
for epoch in range(epochs):
    for batch_samples in train_iterator :
        pass

Luojiaset/net平台登陆异常，忘记密码不能发送验证码

Expected Behavior

Current Behavior

Context

Steps to Reproduce

Your Environment

QuadSearch_GID示例代码中的geobject依赖库缺少

您好，该依赖库并未在github仓库中，且无其他公开渠道获取。
请问该如何获取该库文件？

luojianet并行训练效率提升问题

luojianet团队你们好!
在学习luojianet过程中，我们对分布式并行训练比较感兴趣。我们尝试了在单机多GPU环境下进行遥感影像语义分割的分布式并行训练，使用的模型是UNET语义分割模型。并且进行了一组对比实验，即不使用并行模式，直接在单GPU环境下进行训练。结果发现这两种训练方式耗时相差无几，不知道是什么原因，是因为多卡之间通信也有耗时吗？

还有，luojianet提供的这种并行训练它支持多机多GPU模式吗？

此外，当前Pytorch框架下也配备有分布式并行训练库，不知道在效率，精度方面，luojianet并行与Pytorch并行相比有什么区别。

LuojiaNET如何处理大画幅遥感图片

LuojiaNET团队您好

我从龚健雅院士的《龚健雅院士：遥感大数据与智能解译》报告初次了解到LuojiaNET框架。LuojiaNET针对大画幅遥感图像整体载入进行训练和推理做了框架层面优化。具介绍包括全局和局部影像关联、算子并行两种策略。现在我对模型如何进行大画幅训练和推理依然有不少问题，希望LuojiaNET团队可以给与解答。

• 大画幅遥感图像训练中利用到框架提供的大画幅加载功能，可否提供使用样例作为参考。
• 全局和局部影像关联策略是否只在模型加载大画幅图像时生效，不参与隐藏层大画幅处理。
• 算子并行策略是否是针对特征自动进行分割，类似数据并行。只是一般数据并行为分割Batch，而这里是对特征进行分割。
• 全局和局部影响管理、算子并行策略分别在源代码哪些目录实现。

感激回答。

58.48.42.237/luojianet/ 网站怎么这几天访问不了

58.48.42.237/luojianet/ 网站怎么这几天访问不了呢

是否有适配ubuntu 22.04 & cuda11.2的安装包呢?

如题