export

Function

Reads the sparse table checkpoint saved in the current step and saves it as a NumPy file in key-value format. The key indicates the query ID, and the value indicates the embedding layer. You can use this API to export sparse table data in the following modes:

  • On-chip memory (non-dynamic capacity expansion)
  • DDR mode

Prototype

1
def export(table_list=None):

Parameters

Parameter

Type

Mandatory/Optional

Description

table_list

list[str]

Optional

List of sparse tables to be exported. If no parameter is passed, all sparse tables are exported by default. Range of the list length: [1, 2^31 – 1]

Return Value

  • Success: None
  • Failure: An exception is thrown.

Example

1
2
3
from mx_rec.saver.sparse import export
table_list=["sparse_embedding_table"]
export(table_list=table_list)

See Also

For details about the API call sequence and example, see Porting and Training.