Hi,
I have A100 GPUs. In the documentation it is explicitly stated that cuDNN is used when GPUs are V100. Does this mean that LSTMs performe worse on A100 GPUs? Can this be improved?
No, the docs need to be updated. Thanks for pointing this out.
Hi,
I have A100 GPUs. In the documentation it is explicitly stated that cuDNN is used when GPUs are V100. Does this mean that LSTMs performe worse on A100 GPUs? Can this be improved?
No, the docs need to be updated. Thanks for pointing this out.