I was wondering if it would be possible to use the same trie for different LM (uni-gram vs 3-gram for example) ?
I haven’t tried this out yet but wanted to contact the forum anyway to know if it is possible or not using deepspeech 0.6.1.
I was wondering if it would be possible to use the same trie for different LM (uni-gram vs 3-gram for example) ?
I haven’t tried this out yet but wanted to contact the forum anyway to know if it is possible or not using deepspeech 0.6.1.
As long as all the LMs are built using the same alphabet file, it should be fine. Note that in the upcoming 0.7.0 release the trie file gets embedded with the LM as a single scorer package, so you won’t be able to do this anyway unless you edit the code to support it.