Han Cai

Pytorch has its internal memory management system, which usually occupies a larger memory size than it actually needs. To get the memory decrease on real hardware devices, you need to...

Hi, The code for training start networks has been provided in this repo. For training plain CNNs with specific settings (architectures and hyperparameters), please refer to "run_simple_convnet.py". The code for...

Hi, Thanks for your interest in our work. I am working on preparing the code release. Please subscribe to our [mailing list](https://forms.gle/Z6DNkRidJ1ouxmUk9) to get informed about our updates.

We have supported segmenting everything with Onnxruntime/TensorRT. Please check the following README file: https://github.com/mit-han-lab/efficientvit/blob/master/demo/sam/README.md

Hi johnyang-nv, We measured the latency on the orin platform using the following command: `trtexec --separateProfileRun --iterations=100 --duration=0 --fp16 --onnx=b3_imagenet.onnx` onnx file: [b3_imagenet.onnx](https://drive.google.com/file/d/1_x8nSV3FPqr3eBloTjtPx9AvNShbEXQR/view?usp=sharing) Details are attached below: [01/08/2024-15:39:39] [I] ===...

Hi twmht, Thank you for your interest. We do have this in our plan. If you want to get informed about our updates, please join our mailing list [here](https://forms.gle/Z6DNkRidJ1ouxmUk9) or...

Hello, Thanks for sharing your script and your impressive work! Do you mind if I reference your script in our [README](https://github.com/mit-han-lab/efficientvit#benchmarking-with-tensorrt)? Best, Han

Hi Carlos, I'm delighted to see the integration of EfficientViT-SAM models into your software. Congratulations on the release! Best, Han

Hi mrswang1, The training guide for ImageNet classification is available at https://github.com/mit-han-lab/efficientvit/blob/master/TRAINING.md. Does the issue still exist? If so, can you give me an example for debugging? Thanks, Han

Hi, Can you clarify your question? Do you want to train the model from scratch on CIFAR10 or fine-tune the model on CIFAR10? Thank you, Han