-
Hi~ Great work there!
What I want to ask is whether RepGhost has suffered a serious loss after INT8 quantization?
Or how do you solve quantitative problems? Thanks~
-
tkx for your wonderful work, I found the `timm` is not support shufflenet v1 and v2 yet, there is any plan to support it? and [here](https://github.com/megvii-model/ShuffleNet-Series) is the official …
-
When use load_state_dict(torch.load(/mypath/repghostnet_1_0x_142M_74.22.pth.tar')),It shows Missing key(s) in state_dict:"conv_stem.weight", "bn1.weight", "bn1.bias", "bn1.running_mean", "bn1.running_…
-
您好,请问使用repghostnet进行训练之后,怎么将其转换为推理时的结构,是使用convert.py文件吗?
-
![image](https://github.com/ChengpengChen/RepGhost/assets/52490943/219d74db-65de-4cd3-80c9-f1c4f1642ec6)
1. 请问表格中RepGhostNet 0.5对应的模型是是这个吗https://github.com/ChengpengChen/RepGhost/blob/main/model/rep…
-
Hi, thank you for the great work. I used the pre-trained models to train on my dataset. However, I encountered the following error when running the code.
```
raise RuntimeError('Error(s) in loadin…
-
我测试了你们提供的参数量和Flops计算脚本,发现重参数化前后没有明显区别。但是权重文件明显变小了,可以更新一下计算脚本吗?
```python
from tools import cal_flops_params
input = torch.randn(1, 3, 224, 224)
print("[ Train Model ]")
flops, par…
-
Hi,
I trained the model on my dataset, and then convert it for inference using fille convert.py with the command " python convert.py load TRAINEDMODEL/best_repghost.pkl -m repghost.repghostnet". Howe…
-
hi, any exp result on comparasion with yolov5? Retinanet itself didn't quite fast on mobile platform.
How about mAP compare with yolov5-tiny?