embedding_workspace_calculator

Cal vocabulary_size_per_gpu and workspace_size_per_gpu_in_mb

This tool is used to help you configure proper workspace_size_per_gpu_in_mb required by hugectr.SparseEmbedding. It will calculate least required workspace_size_per_gpu_in_mb based on:

1. slot_size_array: the cardinality array of input features.
2. vvgpu: gpu configuration.
3. emb_vec_size: the embedding vector size.
4. optimizer: optimizer type. Can be adam, adagrad, momentumsgd, nesterov and sgd
5. optimzer update type: optimizer update type. Can be local, global, lazy_global

The default configuration is:

vvgpu = [[0]]
emb_vec_size = 16
optimizer = "adam"
optimizer_update_type = "global"
slot_size_array = [39884, 39043, 17289, 7420, 20263, 3, 7120, 1543, 39884, 39043, 17289, 7420, 20263, 3, 7120, 1543, 63, 63, 39884, 39043, 17289, 7420, 20263, 3, 7120, 1543]

Please refer more details in QAList.md#24