cwq159
commented
3 years ago font{
line-height: 1.6;
}
ul,ol{
padding-left: 20px;
list-style-position: inside;
}
On my GPU with 48GB memory, the timesteps can be 128/256 when the batch size is set to 1.
cwq
***@***.***
签名由
网易邮箱大师
定制
On 07/5/2021 ***@***.***> wrote： Dear author, Thanks for sharing the code. I have trained an ANN on the VOC dataset and got an mAP of 55.2%. However, after I converted the ANN to SNN by ann_to_snn.py, the mAP fell to 31.6%. (On my GPU with 32GB memory, the timesteps cannot be larger than 128 even if the batch size is set to 1.) I wonder how large the GPU memory and the timesteps you used to achieve the mAP of 55.56% on the VOC dataset? I appreciate your reply. Thank you.
—You are receiving this because you are subscribed to this thread.Reply to this email directly, view it on GitHub, or unsubscribe.
Dear author, Thanks for sharing the code. I have trained an ANN on the VOC dataset and got an mAP of 55.2%. However, after I converted the ANN to SNN by ann_to_snn.py, the mAP fell to 31.6%. (On my GPU with 32GB memory, the timesteps cannot be larger than 128 even if the batch size is set to 1.) I wonder how large the GPU memory and the timesteps you used to achieve the mAP of 55.56% on the VOC dataset? I appreciate your reply. Thank you.