This article outlines several Translation Memory (TM) guidelines for making the most of Lilt's systems.
Content guidelines
To ensure the highest quality model adaptation, translation memory files should mostly consist of full natural-language sentences. Uploading a large quantity of short segments may degrade model performance and require a model reset.
Size guidelines
The maximum file size for TM uploads is 200 MB. For file sizes larger than 200 MB, the files can be zipped before upload.
For on-premise customers, processing of TMs larger than 3,000 parallel segments is only recommended when a GPU is installed on the system. Otherwise, memory updates may be slow.
Lilt processes TMs much faster when using a GPU. The exact rate varies by language pair, type of GPU, and amount of CPU and RAM available to the system.