-
### 🐛 Describe the bug
Category | Model | Accuracy
-- | -- | --
timm_models_amp_bf16_training | botnet26t_256 | fail_accuracy
timm_models_amp_fp16_training | botnet26t_…
-
Thanks for your wonderful work!
I prepare to try MobileVit on small dataset, such as MNIST, and I need adjust the network structure. Before this work, I want to know if MobileVit has a better perfo…
-
-
Hi, congratulations on your excellent work!
I would really appreciate if you could help me through this.
So I run
```python
PYTHONWARNINGS="ignore" cvnets-train --common.config-file config/classi…
-
thanks you for the great work;
here is the paper's graph:
![Uploading image.png…]()
I print the layer input and output below:
0 fc x.shape torch.Size([1, 3, 224, 224])
1 fc y.shape torch.Size([…
-
大佬您好,首先感谢您的分享。我运行了一下这份代码,主要跑了mobilevit和mobilenet两个网络,目前有一些问题想问您。
1,直接下载原代码,不改网络结构与参数,得到的精度和readme里的表格有差距。
Mobilevit_s,第160个epoch时,train_acc:0.931(表格中是98.83%);valid_acc:0.906(表格中是92.50%),再继续训练发生过拟合,精…
-
I tried to load a pretrained mobilevit model downloaded from model zoo for fine-tuning on a classification task with 6 labels. I do have excluded the last layer:
![image](https://github.com/apple/ml-…
-
Hello all,
The model training process with customered data was running great with --model_type==mobilevit_s. But I have not figured out this issue when converting the .h5 model into .mlmodel using co…
-
### Feature request
I am trying to train off-line RL using decision transformer, convert to .onnx.
```
from pathlib import Path
from transformers.onnx import FeaturesManager
feature = "seq…
-
**Describe the bug**
estimate model size is different with nvidia-smi usage
**To Reproduce**
1. used code, and command line
2. The code will run on the cuda:2 device
```
import torch
impor…