shreybirmiwal

2 ms·

1. 1.▲
   SparklingTree: 30-40% faster specdec than DSpark by combining DDTree and DSpark(jwlabs.vercel.app)

   2 points by shreybirmiwal 2mo ago|0 comments
2. 2.▲
   by shreybirmiwal 2mo ago

   thanks a lot
3. 3.▲
   LLM speculative decoding explained simply(jwlabs.vercel.app)

   2 points by shreybirmiwal 2mo ago|0 comments
4. 4.▲
   by shreybirmiwal 2mo ago

   thank you so much
5. 5.▲
   Predictive Speculative KV Replication for Bursty LLM Inference(jwlabs.vercel.app)

   41 points by shreybirmiwal 2mo ago|4 comments
6. 6.▲
   Specializing Speculative Decoding to increase drafter accuracy(jwlabs.vercel.app)

   2 points by shreybirmiwal 2mo ago|0 comments