Closed DucNguyenAIML closed 10 months ago
@JackJrs Please provide specific hardware specifications.
In the case of using low-performance GPU, you could try to use PyTorch AMP ( Automatic Mixed Precision) to automatically handle mixed-precision training. AMP aims to improve the training speed and memory utilization while maintaining the desired model accuracy.
@JackJrs Please provide specific hardware specifications.
In the case of using low-performance GPU, you could try to use PyTorch AMP ( Automatic Mixed Precision) to automatically handle mixed-precision training. AMP aims to improve the training speed and memory utilization while maintaining the desired model accuracy.
Hi. Sorry for my late. Here are my hardware: CPU: Intel core i3-6100 CPU @ 3.70GHz (4 CPUs) Graphic card: NVIDIA GeForce GTX 1660
@JackJrs The GTX 1660 is equipped with only 6 GB of GDDR5 memory. Maybe the current code cannot be directly run on this hardware setup.
We run the code with NVIDIA GTX 3090 which has 24 GB video memory.
I run the trainning code and get the error: CUDA out of memory I've already check the Issues discussion in VPG by AndyZeng. However, the code instruction changed too much. Please help.