Closed letdivedeep closed 3 years ago
Hi,
Basically, DetCo is built on MoCo, which cannot be trained on a single GPU due to batch shuffling problem. Please refer to MoCo paper for details.
@shuuchen thanks for the reply can we use two GPU or what the min GPU we should used
docker run -it --runtime=nvidia -e NVIDIA_VISIBLE_DEVICES=0,1 detectron2:v0
You can try two GPUs. Please refer to the updates for GPU configurations.
You can try two GPUs. Please refer to the updates for GPU configurations.
thanks for the help, @shuuchen will try your suggestion.
Hi @shuuchen
I want to train the pre-trained model on the downstream task of object detection. I used the pre-trained model of mocov2 with 800 epochs here
I have followed the following process step 1: Install detectron2.
step 2: Convert a pre-trained MoCo model to detectron2's format:
step 3: Run training:
The only change I did is used a single GPU rather than 8 GPU
I am getting the following error an
how can we run the training on a single GPU? attached are the logs for details log.txt