Training on Colab Pro+ with 16GB GPU RAM and 54GB CPU, which should be enough for most models, but after a 6 epochs, the GPU runs out of memory. I suspect it's a caching issue and not an absolute memory issue. Does anyone know how to clear GPU in between epochs?
Training on Colab Pro+ with 16GB GPU RAM and 54GB CPU, which should be enough for most models, but after a 6 epochs, the GPU runs out of memory. I suspect it's a caching issue and not an absolute memory issue. Does anyone know how to clear GPU in between epochs?