asteroza + learning + gpu   17

Vector and Line Quantization for Billion-scale Similarity Search on GPUs | hgpu.org
Hrm, I wonder if this hierarchical inverted index structure could be applied to specialist database software generally, or perhaps a general GPU database?
similarity  search  image  video  GPU  deep  machine  learning  approximate  nearest  neighbor 
january 2019 by asteroza
NVIDIA/waveglow: A Flow-based Generative Network for Speech Synthesis
25x realtime, so you can respond in realtime if your text generating NN can keep up with the speech-to-text...
Nvidia  GPU  CUDA  deep  machine  learning  human  speech  synthesis  synthetic  voice  artificial  generative  network  software  audio  TTS 
november 2018 by asteroza
yahoo/CaffeOnSpark
Yahoo built a frankenstein framework to support frankenstein hadoop clusters, with a custom MPI over RDMA protocol, infiniband interconnects,and Yarn scheduling SPark and other Deep Learning workloads. The smart approach is an integrated single cluster, so deep learning workloads stay in cluster, but can migrate between GPU and CPU stages.
Caffe  Spark  hadoop  deep  learning  cluster  framework  GPU  GPGPU  CPU  yarn  scheduling  MPI  RDMA  infiniband  yahoo  opensource  software  Delicious 
february 2016 by asteroza

Copy this bookmark:



description:


tags: