snuspl/rnb

Issues

• Investigate the performance effect of `torch.backends.cudnn.benchmark`

  #72 opened 5 years ago by yunseong
  0

• Repo-level README containing general descriptions

  #2 opened 6 years ago by jsjason
  2

• Sanity check on command line arguments

  #6 opened 6 years ago by jsjason
  5

• Pass additional arguments from runner to RunnerModel

  #31 opened 6 years ago by jsjason
  7

• RBPN porting on RnB

  #52 opened 6 years ago by Haabibi
  0

• Return correct output shape for R(2+1)D

  #69 opened 6 years ago by jsjason
  0

• Include GPU path in logs

  #65 opened 6 years ago by jsjason
  0

• Segment-based parallel execution

  #44 opened 6 years ago by jsjason
  0

• Allow input/output of a Runner other than Tensor

  #41 opened 6 years ago by yunseong
  1

• Pass primitive signals instead of PTH tensors to avoid serialization/ipc costs

  #57 opened 6 years ago by jsjason
  0

• Do not block write when a released shared output tensor exists

  #59 opened 6 years ago by jsjason
  0

• NVVL-PyTorch1.1 hangs when program exits

  #56 opened 6 years ago by jsjason
  0

• Decouple dataset-specific part from client implementation

  #40 opened 6 years ago by yunseong
  0

• Per-GPU queue for aggressive pipelining

  #43 opened 6 years ago by jsjason
  0

• Implement a no-pipelining version of R(2+1)D

  #42 opened 6 years ago by jsjason
  0

• Link with CUPTI to acquire occupancy metrics

  #19 opened 6 years ago by jsjason
  0

• Decouple loader implementation from our main logic

  #37 opened 6 years ago by yunseong
  2

• Visualize the progress of inferences

  #38 opened 6 years ago by yunseong
  0

• Take mean_interval = 0 as a special case

  #12 opened 6 years ago by yunseong
  0

• Monitor GPU memory usage

  #33 opened 6 years ago by yunseong
  0

• Let Neural Network be split

  #13 opened 6 years ago by yunseong
  0

• Clean up and encapsulate multiprocessing semaphore logic

  #5 opened 6 years ago by jsjason
  2

• Support heterogeneous placement

  #14 opened 6 years ago by yunseong
  0

• Change the way to express the constraints of command-line arguments

  #23 opened 6 years ago by yunseong
  1

• Decouple neural network implementation from runner logic

  #20 opened 6 years ago by jsjason
  0

• Subtle behavior of python IPC

  #17 opened 6 years ago by yunseong
  1

• Loader finishes earlier than runners

  #7 opened 6 years ago by jsjason
  1

• Periodically log queue size

  #18 opened 6 years ago by jsjason
  0

• Multiple loaders

  #4 opened 6 years ago by jsjason
  0

• Update PyTorch version

  #3 opened 6 years ago by jsjason
  0