rpdaiml

21 ms·

1. 1.▲
   by rpdaiml 5mo ago

   Nice work, that throughput is wild.
2. 2.▲
   by rpdaiml 5mo ago

   How are you handling the on device speech pipeline, especially around model size, latency, and accuracy tradeoffs on consumer hardware?
3. 3.▲
   by rpdaiml 6mo ago

   This is a nice idea. A tiny implementation can be way more useful for learning than yet another wrapper around a big model, especially if it keeps the training loop and inference path small enough to read end to end.
4. 4.▲
   by rpdaiml 6mo ago

   curious and wondering its porting and usage on edge devices?