Why does enabling the choosing of a seed result in out of memory errors?

[G-force78]

I added the option to choose a seed between 1-2000, this caused out of memory error every single time.

              PyTorch CUDA memory summary, device ID 0                 |

---------------------------------------------------------------------------		CUDA OOMs: 0	cudaMalloc retries: 0		===========================================================================		Metric	Cur Usage	Peak Usage	Tot Alloc	Tot Freed
Allocated memory	3435 MiB	3435 MiB	3520 MiB	87264 KiB
from large pool	3295 MiB	3295 MiB	3317 MiB	23040 KiB
from small pool	140 MiB	200 MiB	202 MiB	64224 KiB
---------------------------------------------------------------------------
Active memory	3435 MiB	3435 MiB	3520 MiB	87264 KiB
from large pool	3295 MiB	3295 MiB	3317 MiB	23040 KiB
from small pool	140 MiB	200 MiB	202 MiB	64224 KiB
---------------------------------------------------------------------------
Requested memory	3395 MiB	3395 MiB	3480 MiB	87264 KiB
from large pool	3255 MiB	3255 MiB	3278 MiB	23040 KiB
from small pool	139 MiB	199 MiB	202 MiB	64224 KiB
---------------------------------------------------------------------------
GPU reserved memory	3608 MiB	3608 MiB	3608 MiB	0 B
from large pool	3404 MiB	3404 MiB	3404 MiB	0 B
from small pool	204 MiB	204 MiB	204 MiB	0 B
---------------------------------------------------------------------------
Non-releasable memory	119582 KiB	142451 KiB	1551 MiB	1435 MiB
from large pool	111426 KiB	133186 KiB	1341 MiB	1232 MiB
from small pool	8156 KiB	33618 KiB	210 MiB	202 MiB
---------------------------------------------------------------------------
Allocations	2002	2002	2290	288
from large pool	510	510	522	12
from small pool	1492	1492	1768	276
---------------------------------------------------------------------------
Active allocs	2002	2002	2290	288
from large pool	510	510	522	12
from small pool	1492	1492	1768	276
---------------------------------------------------------------------------
GPU reserved segments	263	263	263	0
from large pool	161	161	161	0
from small pool	102	102	102	0
---------------------------------------------------------------------------
Non-releasable allocs	124	138	300	176
from large pool	73	82	156	83
from small pool	51	85	144	93
---------------------------------------------------------------------------
Oversize allocations	0	0	0	0
---------------------------------------------------------------------------
Oversize GPU segments	0	0	0	0