Feature Request: Being able to only reload the graph
- Vorherrschende Sprache
- Jupyter Notebook
- Sterne
- 15.1k
- Forks
- 1.8k
- PR-Merge-Kennzahlen
- Keine gemergten PRs in 30 T.
Beschreibung
First of all, thanks for publishing this nice API. I really appreciate the well-thought-out API-design and hope I can contribute to it here and there.
To the topic: Let's say I have built certain domain specific graphs or graphs with an extended vocabulary or virtually anything language model specific. In these cases it would be nice to be able to reuse the already loaded acoustic model but only load the graph specific stuff as adjusted Model instance. This could for example be realized by adding something like `Model loadGraph(const char *graph_path);` or `Model loadGraph(const char *graph_path, const char *rescore_path);` to the Model class. Especially in a server environment where potentially lots of models / graphs are preloaded, this feature would be very useful in terms of memory footprint.
How do you think about that?
Beitragsleitfaden
Für dieses Repository ist kein Beitragsleitfaden indexiert
Bewertung
Dieses Issue wurde noch nicht bewertet.