Our tests showed that the tuner is currently making the wrong decision at 4GB on 16 P5s, which caused a regression. This is a temporary workaround to force NVLSTREE while we work to make the model more accurate.
By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.
Our tests showed that the tuner is currently making the wrong decision at 4GB on 16 P5s, which caused a regression. This is a temporary workaround to force NVLSTREE while we work to make the model more accurate.
By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.