* Check if file exist. */

* Get a file name from the list. * @return the file name and id. */

if we need more configuration about EmbeddingCollection

Initializes the MLP layer. Args: units: Sequential list of layer sizes. List[int]. use_bias: Whether to include a bias term.

generate parquet files and return data in the form of `row-major` stored in vector<>

This is a heuristic based on greedy policy. The shard is placed to the bucket with the lowest hotness cost

Initializes the MLP layer. Args: units: Sequential list of layer sizes. List[int]. use_bias: Whether to include a bias term.

Initializes the MLP layer. Args: units: Sequential list of layer sizes. use_bias: Whether to include a bias term. acti

Initializes the MLP layer. Args: units: Sequential list of layer sizes. use_bias: Whether to include a bias term. acti

Abbreviated as ``sok.assign``. Assign the indices and value tensor to the target variable. Parameters ---------- var: sok.Dynam

Abbreviated as ``sok.dump``. Dump the embedding tables into one folder , will generate key and value file per table. Table name prefix