We've just released new improvements!
OctoML now supports acceleration and packaging on large models up to 78GB, such as DLRM
If you would like to run workflows on the order of gigabytes, let your customer account representative know to access our custom high memory T4 and V100 instances on GCP
We've upgraded our TensorFlow, TFLite, and Keras support to versions 2.0-2.6
We previously supported versions 2.0-2.2