显存估算精度变差
@wengjun 遇到了爆显存问题:
juno/OSIRIS/production#39 (comment 40928)
我也遇到了:
╭─heroxbd@cat /mnt/stage/OSIRIS/calibration/FSMP ‹master›
╰─$ ipython3 -i ./fsmp.py -- sparsify/0310/003616/0.h5 -o fsmp/0310/003616/0.pq --ref SER.h5
File /opt/gentoo/usr/lib/python3.11/site-packages/cupy/_creation/from_data.py:75, in asarray(a, dtype, order)
49 def asarray(a, dtype=None, order=None):
50 """Converts an object to array.
51
52 This is equivalent to ``array(a, dtype, copy=False, order=order)``.
(...)
73
74 """
---> 75 return _core.array(a, dtype, False, order)
File cupy/_core/core.pyx:2376, in cupy._core.core.array()
File cupy/_core/core.pyx:2400, in cupy._core.core.array()
File cupy/_core/core.pyx:2547, in cupy._core.core._array_default()
File cupy/cuda/memory.pyx:483, in cupy.cuda.memory.MemoryPointer.copy_from_host_async()
File cupy_backends/cuda/api/runtime.pyx:935, in cupy_backends.cuda.api.runtime.streamIsCapturing()
File cupy_backends/cuda/api/runtime.pyx:144, in cupy_backends.cuda.api.runtime.check_status()
CUDARuntimeError: cudaErrorIllegalAddress: an illegal memory access was encountered