哈哈，多谢提醒

## 所有的在线运行文档 字数限制，还有一部分有在线运行标签的文档罗列情况： |序号|文档名称|ModelArts连接| |:-----|:-----|:-----| |30|[Parameter](https://www.mindspore.cn/docs/programming_guide/zh-CN/master/parameter_introduction.html)|[](https://authoring-modelarts-cnnorth4.huaweicloud.com/console/lab?share-url-b64=aHR0cHM6Ly9vYnMuZHVhbHN0YWNrLmNuLW5vcnRoLTQubXlodWF3ZWljbG91ZC5jb20vbWluZHNwb3JlLXdlYnNpdGUvbm90ZWJvb2svbWFzdGVyL3Byb2dyYW1taW5nX2d1aWRlL3poX2NuL21pbmRzcG9yZV9wYXJhbWV0ZXJfaW50cm9kdWN0aW9uLmlweW5i&imageid=65f636a0-56cf-49df-b941-7d2a07ba8c8c) |31|[通用数据处理](https://www.mindspore.cn/docs/programming_guide/zh-CN/master/pipeline_common.html)|[](https://authoring-modelarts-cnnorth4.huaweicloud.com/console/lab?share-url-b64=aHR0cHM6Ly9vYnMuZHVhbHN0YWNrLmNuLW5vcnRoLTQubXlodWF3ZWljbG91ZC5jb20vbWluZHNwb3JlLXdlYnNpdGUvbm90ZWJvb2svbW9kZWxhcnRzL3Byb2dyYW1taW5nX2d1aWRlL21pbmRzcG9yZV9waXBlbGluZS5pcHluYg==&imageid=65f636a0-56cf-49df-b941-7d2a07ba8c8c) |32|[运行方式](https://www.mindspore.cn/docs/programming_guide/zh-CN/master/run.html)|[](https://authoring-modelarts-cnnorth4.huaweicloud.com/console/lab?share-url-b64=aHR0cHM6Ly9vYnMuZHVhbHN0YWNrLmNuLW5vcnRoLTQubXlodWF3ZWljbG91ZC5jb20vbWluZHNwb3JlLXdlYnNpdGUvbm90ZWJvb2svbW9kZWxhcnRzL3Byb2dyYW1taW5nX2d1aWRlL21pbmRzcG9yZV9ydW4uaXB5bmI=&imageid=65f636a0-56cf-49df-b941-7d2a07ba8c8c) |33|[数据采样](https://www.mindspore.cn/docs/programming_guide/zh-CN/master/sampler.html)|[](https://authoring-modelarts-cnnorth4.huaweicloud.com/console/lab?share-url-b64=aHR0cHM6Ly9vYnMuZHVhbHN0YWNrLmNuLW5vcnRoLTQubXlodWF3ZWljbG91ZC5jb20vbWluZHNwb3JlLXdlYnNpdGUvbm90ZWJvb2svbW9kZWxhcnRzL3Byb2dyYW1taW5nX2d1aWRlL21pbmRzcG9yZV9zYW1wbGVyLmlweW5i&imageid=65f636a0-56cf-49df-b941-7d2a07ba8c8c) |34|[保存模型](https://www.mindspore.cn/docs/programming_guide/zh-CN/master/save_model.html)|[](https://authoring-modelarts-cnnorth4.huaweicloud.com/console/lab?share-url-b64=aHR0cHM6Ly9vYnMuZHVhbHN0YWNrLmNuLW5vcnRoLTQubXlodWF3ZWljbG91ZC5jb20vbWluZHNwb3JlLXdlYnNpdGUvbm90ZWJvb2svbW9kZWxhcnRzL21pbmRzcG9yZV9zYXZlX21vZGVsLmlweW5i&imageid=65f636a0-56cf-49df-b941-7d2a07ba8c8c) |35|[Tensor](https://www.mindspore.cn/docs/programming_guide/zh-CN/master/tensor.html)|[](https://authoring-modelarts-cnnorth4.huaweicloud.com/console/lab?share-url-b64=aHR0cHM6Ly9vYnMuZHVhbHN0YWNrLmNuLW5vcnRoLTQubXlodWF3ZWljbG91ZC5jb20vbWluZHNwb3JlLXdlYnNpdGUvbm90ZWJvb2svbW9kZWxhcnRzL3Byb2dyYW1taW5nX2d1aWRlL21pbmRzcG9yZV90ZW5zb3IuaXB5bmI=&imageid=65f636a0-56cf-49df-b941-7d2a07ba8c8c) |36|[文本处理与增强](https://www.mindspore.cn/docs/programming_guide/zh-CN/master/tokenizer.html)|[](https://authoring-modelarts-cnnorth4.huaweicloud.com/console/lab?share-url-b64=aHR0cHM6Ly9vYnMuZHVhbHN0YWNrLmNuLW5vcnRoLTQubXlodWF3ZWljbG91ZC5jb20vbWluZHNwb3JlLXdlYnNpdGUvbm90ZWJvb2svbW9kZWxhcnRzL3Byb2dyYW1taW5nX2d1aWRlL21pbmRzcG9yZV90b2tlbml6ZXIuaXB5bmI=&imageid=65f636a0-56cf-49df-b941-7d2a07ba8c8c) |37|[自动微分](https://www.mindspore.cn/tutorials/zh-CN/master/autograd.html)|[](https://authoring-modelarts-cnnorth4.huaweicloud.com/console/lab?share-url-b64=aHR0cHM6Ly9taW5kc3BvcmUtd2Vic2l0ZS5vYnMuY24tbm9ydGgtNC5teWh1YXdlaWNsb3VkLmNvbS9ub3RlYm9vay9tYXN0ZXIvdHV0b3JpYWxzL3poX2NuL21pbmRzcG9yZV9hdXRvZ3JhZC5pcHluYg==&imageid=65f636a0-56cf-49df-b941-7d2a07ba8c8c) |38|[数据加载及处理](https://www.mindspore.cn/tutorials/zh-CN/master/dataset.html)|[](https://authoring-modelarts-cnnorth4.huaweicloud.com/console/lab?share-url-b64=aHR0cHM6Ly9taW5kc3BvcmUtd2Vic2l0ZS5vYnMuY24tbm9ydGgtNC5teWh1YXdlaWNsb3VkLmNvbS9ub3RlYm9vay9tb2RlbGFydHMvcXVpY2tfc3RhcnQvbWluZHNwb3JlX2RhdGFzZXQuaXB5bmI=&imageid=65f636a0-56cf-49df-b941-7d2a07ba8c8c) |39|[简单线性函数拟合](https://www.mindspore.cn/tutorials/zh-CN/master/linear_regression.html)|[](https://authoring-modelarts-cnnorth4.huaweicloud.com/console/lab?share-url-b64=aHR0cHM6Ly9taW5kc3BvcmUtd2Vic2l0ZS5vYnMuY24tbm9ydGgtNC5teWh1YXdlaWNsb3VkLmNvbS9ub3RlYm9vay9tb2RlbGFydHMvbWluZHNwb3JlX2xpbmVhcl9yZWdyZXNzaW9uLmlweW5i&imageid=65f636a0-56cf-49df-b941-7d2a07ba8c8c) |40|[建立神经网络](https://www.mindspore.cn/tutorials/zh-CN/master/model.html)|[](https://authoring-modelarts-cnnorth4.huaweicloud.com/console/lab?share-url-b64=aHR0cHM6Ly9taW5kc3BvcmUtd2Vic2l0ZS5vYnMuY24tbm9ydGgtNC5teWh1YXdlaWNsb3VkLmNvbS9ub3RlYm9vay9tYXN0ZXIvdHV0b3JpYWxzL3poX2NuL21pbmRzcG9yZV9tb2RlbC5pcHluYg==&imageid=65f636a0-56cf-49df-b941-7d2a07ba8c8c) |41|[优化模型参数](https://www.mindspore.cn/tutorials/zh-CN/master/optimization.html)|[](https://authoring-modelarts-cnnorth4.huaweicloud.com/console/lab?share-url-b64=aHR0cHM6Ly9taW5kc3BvcmUtd2Vic2l0ZS5vYnMuY24tbm9ydGgtNC5teWh1YXdlaWNsb3VkLmNvbS9ub3RlYm9vay9tb2RlbGFydHMvcXVpY2tfc3RhcnQvbWluZHNwb3JlX29wdGltaXphdGlvbi5pcHluYg==&imageid=65f636a0-56cf-49df-b941-7d2a07ba8c8c) |42|[初学入门](https://www.mindspore.cn/tutorials/zh-CN/master/quick_start.html)|[](https://authoring-modelarts-cnnorth4.huaweicloud.com/console/lab?share-url-b64=aHR0cHM6Ly9taW5kc3BvcmUtd2Vic2l0ZS5vYnMuY24tbm9ydGgtNC5teWh1YXdlaWNsb3VkLmNvbS9ub3RlYm9vay9tYXN0ZXIvdHV0b3JpYWxzL3poX2NuL21pbmRzcG9yZV9xdWlja19zdGFydC5pcHluYg==&imageid=65f636a0-56cf-49df-b941-7d2a07ba8c8c) |43|[张量](https://www.mindspore.cn/tutorials/zh-CN/master/tensor.html)|[](https://authoring-modelarts-cnnorth4.huaweicloud.com/console/lab?share-url-b64=aHR0cHM6Ly9taW5kc3BvcmUtd2Vic2l0ZS5vYnMuY24tbm9ydGgtNC5teWh1YXdlaWNsb3VkLmNvbS9ub3RlYm9vay9tYXN0ZXIvdHV0b3JpYWxzL3poX2NuL21pbmRzcG9yZV90ZW5zb3IuaXB5bmI=&imageid=65f636a0-56cf-49df-b941-7d2a07ba8c8c) |44|[高级数据集管理](https://www.mindspore.cn/tutorials/zh-CN/master/intermediate/data.html)|[](https://authoring-modelarts-cnnorth4.huaweicloud.com/console/lab?share-url-b64=aHR0cHM6Ly9taW5kc3BvcmUtd2Vic2l0ZS5vYnMuY24tbm9ydGgtNC5teWh1YXdlaWNsb3VkLmNvbS9ub3RlYm9vay9tYXN0ZXIvdHV0b3JpYWxzL3poX2NuL21pbmRzcG9yZV9kYXRhLmlweW5i&imageid=65f636a0-56cf-49df-b941-7d2a07ba8c8c) |45|[动态图与静态图](https://www.mindspore.cn/tutorials/zh-CN/master/intermediate/pynative_mode_and_graph_mode.html)|[](https://authoring-modelarts-cnnorth4.huaweicloud.com/console/lab?share-url-b64=aHR0cHM6Ly9taW5kc3BvcmUtd2Vic2l0ZS5vYnMuY24tbm9ydGgtNC5teWh1YXdlaWNsb3VkLmNvbS9ub3RlYm9vay9yMS41L3R1dG9yaWFscy96aF9jbi9taW5kc3BvcmVfcHluYXRpdmVfbW9kZV9hbmRfZ3JhcGhfbW9kZS5pcHluYg==&imageid=59a6e9f5-93c0-44dd-85b0-82f390c5d53b) |46|[对抗示例生成](https://www.mindspore.cn/tutorials/zh-CN/master/intermediate/image_and_video/adversarial_example_generation.html)|[](https://authoring-modelarts-cnnorth4.huaweicloud.com/console/lab?share-url-b64=aHR0cHM6Ly9taW5kc3BvcmUtd2Vic2l0ZS5vYnMuY24tbm9ydGgtNC5teWh1YXdlaWNsb3VkLmNvbS9ub3RlYm9vay9yMS41L3R1dG9yaWFscy96aF9jbi9taW5kc3BvcmVfYWR2ZXJzYXJpYWxfZXhhbXBsZV9nZW5lcmF0aW9uLmlweW5i&imageid=59a6e9f5-93c0-44dd-85b0-82f390c5d53b) |47|[深度卷积对抗生成网络](https://www.mindspore.cn/tutorials/zh-CN/master/intermediate/image_and_video/dcgan.html)|[](https://authoring-modelarts-cnnorth4.huaweicloud.com/console/lab?share-url-b64=aHR0cHM6Ly9taW5kc3BvcmUtd2Vic2l0ZS5vYnMuY24tbm9ydGgtNC5teWh1YXdlaWNsb3VkLmNvbS9ub3RlYm9vay9yMS41L3R1dG9yaWFscy96aF9jbi9taW5kc3BvcmVfZGNnYW4uaXB5bmI=&imageid=59a6e9f5-93c0-44dd-85b0-82f390c5d53b) |48|[Ascend910与GPU推理](https://www.mindspore.cn/tutorials/zh-CN/master/intermediate/inference/ascend910_and_gpu_inference.html)|[](https://authoring-modelarts-cnnorth4.huaweicloud.com/console/lab?share-url-b64=aHR0cHM6Ly9taW5kc3BvcmUtd2Vic2l0ZS5vYnMuY24tbm9ydGgtNC5teWh1YXdlaWNsb3VkLmNvbS9ub3RlYm9vay9tYXN0ZXIvdHV0b3JpYWxzL3poX2NuL21pbmRzcG9yZV9hc2NlbmQ5MTBfYW5kX2dwdV9pbmZlcmVuY2UuaXB5bmI=&imageid=65f636a0-56cf-49df-b941-7d2a07ba8c8c) |49|[使用字符级RNN分类名称](https://www.mindspore.cn/tutorials/zh-CN/master/intermediate/text/rnn_classification.html)|[](https://authoring-modelarts-cnnorth4.huaweicloud.com/console/lab?share-url-b64=aHR0cHM6Ly9taW5kc3BvcmUtd2Vic2l0ZS5vYnMuY24tbm9ydGgtNC5teWh1YXdlaWNsb3VkLmNvbS9ub3RlYm9vay9yMS41L3R1dG9yaWFscy96aF9jbi9taW5kc3BvcmVfcm5uX2NsYXNzaWZpY2F0aW9uLmlweW5i&imageid=59a6e9f5-93c0-44dd-85b0-82f390c5d53b) |50|[使用字符级RNN生成名称](https://www.mindspore.cn/tutorials/zh-CN/master/intermediate/text/rnn_generation.html)|[](https://authoring-modelarts-cnnorth4.huaweicloud.com/console/lab?share-url-b64=aHR0cHM6Ly9taW5kc3BvcmUtd2Vic2l0ZS5vYnMuY24tbm9ydGgtNC5teWh1YXdlaWNsb3VkLmNvbS9ub3RlYm9vay9yMS41L3R1dG9yaWFscy96aF9jbi9taW5kc3BvcmVfcm5uX2dlbmVyYXRpb24uaXB5bmI=&imageid=59a6e9f5-93c0-44dd-85b0-82f390c5d53b) |51|[使用SentimentNet实现情感分类](https://www.mindspore.cn/tutorials/zh-CN/master/intermediate/text/sentimentnet.html)|[](https://authoring-modelarts-cnnorth4.huaweicloud.com/console/lab?share-url-b64=aHR0cHM6Ly9taW5kc3BvcmUtd2Vic2l0ZS5vYnMuY24tbm9ydGgtNC5teWh1YXdlaWNsb3VkLmNvbS9ub3RlYm9vay9tYXN0ZXIvdHV0b3JpYWxzL3poX2NuL21pbmRzcG9yZV9zZW50aW1lbnRuZXQuaXB5bmI=&imageid=65f636a0-56cf-49df-b941-7d2a07ba8c8c) > 另外有个[对抗生成网络](https://www.mindspore.cn/tutorials/zh-CN/master/intermediate/image_and_video/dcgan.html)的在线运行，连接到Modelarts的device居然是华为自研Ascend 910，也可以免费使用，跑起来的速度确实比GPU速度更快一些。但只能使用Modelarts默认适配的MindSpore。不支持安装最新的MindSpore版本

感谢大佬的分享，太详细了，收藏好，以防万一，以后自己编译，可以避免不少大坑

您好，mindir模型有个分界线就是1.1.0，现在的1.1分支的代码已经是1.1.1版本了，这个版本之前的mindir需要用官网发布的1.1.0的convert进行convert。或者mindir用1.1分支的代码重新导出，那么也可以convert

评论 rmzk_00：这样啊，那还没有处理过这样的，等大神来解答吧

如果你的图片是jpg格式的或者jpeg格式的，将训练用的猫图片和狗图片分成两个文件夹0和1存放。比如：放在/home/train/0/ 放小猫图片。 /home/train/1/放小狗图片。下面的代码你直接调用create_dataset("/home/train/")就可以转换成训练用的数据了。import mindspore.dataset as dsimport mindspore.common.dtype as mstypeimport mindspore.dataset.vision.c_transforms as Cimport mindspore.dataset.transforms.c_transforms as C2def create_dataset(datasets_path):    ds1 = ds.ImageFolderDataset(datasets_path, num_parallel_workers=8, extensions=[".jpg", ".JPEG"])    batch_size = 32    image_size = 32    mean = [0.485 * 32, 0.456 * 32, 0.406 * 32]    std = [0.229 * 32, 0.224 * 32, 0.225 * 32]    trans = [                # 图片拉伸，缩放等操作，自己按照需要操作                C.RandomCropDecodeResize(image_size, scale=(0.08, 1.0), ratio=(0.75, 1.333)),                # 随机裁剪图片，这里的0.5表示图片视野的一半                C.RandomHorizontalFlip(prob=0.5),                # 归一化标准化，数值需按照自己的图片实际情况设置                C.Normalize(mean=mean, std=std),                # 通道转换                C.HWC2CHW()    ]    type_cast_op = C2.TypeCast(mstype.int32)    ds1 = ds1.map(input_columns="image", num_parallel_workers=8, operations=trans)    ds1 = ds1.map(input_columns="label", num_parallel_workers=8, operations=type_cast_op)    # shuffle混洗操作必须的    ds1 = ds1.shuffle(buffer_size=100)    # batch操作看内存大小，自己随意设置    ds1 = ds1.batch(batch_size, drop_remainder=True)    # repeat操作数据重复几倍，数字按照自己需要填    ds1 = ds1.repeat(5)        return ds1

链接在文中已经调整成可显示的。

GPU算子开发里的.h文件里边，重写了GetInputSizeList()函数，返回值是一个向量input_size_list_，向量里的元素是每个入参需要的内存大小，这个需要自己计算，看看是不是计算错了

学习率是不是设置过大？

已经提issue确认在master分支解决。