nni
nni copied to clipboard
microsoft
An open source AutoML toolkit for automate machine learning lifecycle, including feature engineering, neural architecture search, model compression and hyper-parameter tuning.
### Description ### fix #4875 - [x] fix infinite loop in `ChannelDependency._get_parent_layers` - [ ] fix buffer can not be traced? - https://github.com/facebookresearch/detectron2/blob/45b3fcea6e76bf7a351e54e01c7d6e1a3a0100a5/detectron2/modeling/anchor_generator.py#L177 - the following `view` in `base_anchors.view(1, -1,...
I checked the doc and some codes to find that many pruners only suport conv2d. Is there any way to prune 3D model? It seems that only fine-grained pruning (Like...
我在使用NNI进行剪枝过程中（L1FilterPruner策略），在m_speedup.speedup_model()这一步产生了报错 TypeError: forward() missing 1 required positional argument: 'x' 我在整个yolov2-tiny中仅对最后一层使用了yolo检测层,需要剪枝部分如下： +-------+----------------------+--------+-------------------+----------+---------+ | Index | Name | Type | Weight Shape | FLOPs | #Params | +-------+----------------------+--------+-------------------+----------+---------+ | 0 | model.0.conv.module...
**Describe the issue**:  It seems that the operation of `aten::expand_as` is still not supported. If possible, please fix this issue as soon. Thanks! @sverrejoh @bgianfo
I cannot view old experiments. When I use the nnictl view command I get an attribute error: raise AttributeError(f'{class_name} does not have field(s) {fields}') Entire error pasted below. (.venv) root@anesth-lars:/algorithm_dev/NeuralNetV1#...
**Describe the issue**: When trying to run an experiment on Azure Machine Learning (AML) an error occurs inside the training service. Error occurs when ```this.trainingService.run()``` is called in nnimanager.js. Unfortunatelly...
**Describe the issue**: During computing Channel Dependencies `reshape_break_channel_dependency` does following code to ensure that the number of input channels equals the number of output channels: ``` in_shape = op_node.auxiliary['in_shape'] out_shape...
**Describe the issue**: As title described, I ran ProxylessNAS example with slight network parameter modification. After 5 epochs, acc 1, acc 5 looks like the same as what I got...
**Describe the issue**: It started out working properly, and then suddenly the problem arose，But the first trail is still functioning   **Environment**: - NNI version:2.6.1 - Training service (local|remote|pai|aml|etc):remote...
When working with task_genrator for torch.jit.ScriptModule, I found that Model Compression uses pickle to model serialization, but this feature is not supported yet because ScriptModule doesn't have the `__getstate__` method....
