mfalt

using CUDA

import CUDA.CUDNN: cudnnConvolutionForward

const W1 = cu(randn(5,5,3,6))

function inference(imgs)
    out = cudnnConvolutionForward(W1, imgs)
    return maximum(Array(out))
end

mfalt

+-----------------------------------------------------------------------------+
| NVIDIA-SMI 470.103.01   Driver Version: 470.103.01   CUDA Version: 11.4     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  NVIDIA GeForce ...  Off  | 00000000:01:00.0 Off |                  N/A |
|  0%   43C    P8    10W / 250W |   2256MiB / 11175MiB |      0%      Default |
|                               |                      |                  N/A |

mfalt

ERROR: LoadError: TaskFailedException
Stacktrace:
 [1] wait
   @ ./task.jl:334 [inlined]
 [2] fetch(t::Task)
   @ Base ./task.jl:349
 [3] top-level scope
   @ ~/knightvision/piece_recognition/KnightVisionServer/test/test_cuda.jl:46

    nested task error: Out of GPU memory trying to allocate 13.184 MiB

mfalt

const LOCK = ReentrantLock()

function inference(imgs)
    lock(LOCK)
    try
        out = nn(imgs)
        return maximum(cpu(out))
    finally
        unlock(LOCK)

mfalt

ERROR: LoadError: TaskFailedException
Stacktrace:
 [1] wait
   @ ./task.jl:334 [inlined]
 [2] fetch(t::Task)
   @ Base ./task.jl:349
 [3] top-level scope
   @ ~/knightvision/piece_recognition/KnightVisionServer/test/test_cuda.jl:39

    nested task error: CUDNNError: CUDNN_STATUS_INTERNAL_ERROR (code 4)

mfalt

# After execution gets stuck, sending `schedule(t, nothing, error=true)` to the task: 
Stacktrace:
  [1] try_yieldto(undo::typeof(Base.ensure_rescheduled))
    @ Base ./task.jl:777
  [2] wait()
    @ Base ./task.jl:837
  [3] wait(c::Base.GenericCondition{ReentrantLock})
    @ Base ./condition.jl:123
  [4] wait(e::Base.Event)
    @ Base ./lock.jl:366

mfalt

ERROR: LoadError: TaskFailedException
Stacktrace:
 [1] wait
   @ ./task.jl:334 [inlined]
 [2] fetch(t::Task)
   @ Base ./task.jl:349
 [3] top-level scope
   @ ~/knightvision/piece_recognition/KnightVisionServer/test/test_cuda.jl:39

    nested task error: CUDNNError: CUDNN_STATUS_INTERNAL_ERROR (code 4)

mfalt

ERROR: LoadError: TaskFailedException

    nested task error: Out of GPU memory trying to allocate 897.217 MiB
    Effective GPU memory usage: 52.47% (5.726 GiB/10.913 GiB)
    Memory pool usage: 3.117 GiB (3.656 GiB reserved)
    Stacktrace:
      [1] macro expansion
        @ ~/.julia/packages/CUDA/0IDh2/src/pool.jl:219 [inlined]
      [2] macro expansion
        @ ./timing.jl:299 [inlined]

mfalt

using Flux

const nn =  gpu(Chain(
            Conv((5, 5), 3=>6, relu),
            MaxPool((2, 2)),
            Conv((5, 5), 6=>16, relu),
            MaxPool((2, 2)),
            flatten,
            Dense(256, 120, relu), 
            Dense(120, 84, relu), 

mfalt

hello