How many computing resources did you use to train the model?

[liuxin-a]

I would like to know if your model can be trained on Nvidia RTX 3090s?

[ShihaoZhaoZSH]

Certainly, it can be trained on NVIDIA RTX 3090s. The batch size have to be adjusted accordingly.

[songjiechong]

I would like to know the GPU numbers while training the model. I train a model with 2 Nvidia V100 and batch size = 1, but it still has the error "CUDA out of memory".

[ShihaoZhaoZSH]

I would like to know the GPU numbers while training the model. I train a model with 2 Nvidia V100 and batch size = 1, but it still has the error "CUDA out of memory".

Please note that the batch size here is for a batch on a single GPU. A V100 with 24GB memory can handle the training with a batch size of 1.

[songjiechong]

I would like to know the GPU numbers while training the model. I train a model with 2 Nvidia V100 and batch size = 1, but it still has the error "CUDA out of memory".

Please note that the batch size here is for a batch on a single GPU. A V100 with 24GB memory can handle the training with a batch size of 1.

Thank you very much.

[liuxin-a]

Certainly, it can be trained on NVIDIA RTX 3090s. The batch size have to be adjusted accordingly. Thank you so much,I would like to know how many gpu hours it take you to train the model and which kind of GPU you used？Looking forward to your reply.