MagicSource

updated it.

lazyconfig only supports MaskRCNN

you should using `train_visdron.py` to train if you register in that file. I have updated the custom dataset training entrance, you can directly edit your dataset and resgistered here: https://github.com/jinfagang/yolov7/blob/main/train_custom_datasets.py...

可以根据gpu num调整一下lr，现在最新版的d2，会默认按照八卡进行配置，也就是说，如果lr设置的0.008，而你训练只有一张卡，d2会自动把lr变成 0.001

@LuoxiaofeiFly 你的确发现了一个BUG，可以给主仓库提交一个PR修正以下这个地方么？看起来似乎不影响运行，但在onnx export的时候恐容易崩

We didn't fully opensource convnext-yolox yet.

没有任何改进，如果你看到训练效果更好，那可能是由于我们的实现与原版的baias造成的。 具体来将： - 模型没有任何改变（当然在yolov7海量backbone包括VIT你可以随意组合）； - learning scheduler可能又些许差异，因为d2里面是按照iteration来算的； - augmentation可能存在差异，因为我们用的是我们的mosiac版本； BTW，我们提供了一个YOLOX-Keypoint 端到端的模型。感兴趣可以莫改一波。可以post一些结果看看差异在哪儿，我们分析一波

@JackeyLovers can u help contribute some of your training result in [discussion](https://github.com/jinfagang/yolov7/discussions) ? So that I can make it as a reference for others who want using yolov7 as training...

it was acceptable, there might be some slightly difference between floats, but it's OK in inference.

@ayoolaolafenwa Hi, thank u for your interest. Theoretically it should able to export to onnx. What's error messages did got when export SparseInst?