What’s the status of Token/sequence limits on NLP tasks as of riva 1.10.0?
Has the Q/A sequence limit been raised?
Release notes don’t list it as a known limitation.
What’s the status of Token/sequence limits on NLP tasks as of riva 1.10.0?
Has the Q/A sequence limit been raised?
Release notes don’t list it as a known limitation.
Would appreciate some updates on this
Hi @ShantanuNair ,
Thanks for your interest in Riva,
Apologies for the delay in response,
Most of our NLP models are bert based, we support a token limit of upto 512, any higher than that would be truncated to first 384 tokens