Memory (Vram) full before completing the process

[Shaddy-ElDesouky]

I have an RTX3000 with 6GB memory, before running the command NVTop(HTOP like for Nvidia) indicates that I have more than 5.5GB free but running the command fills that up, I used the suggested "PYTORCH_CUDA_ALLOC_CONF=max_split_size_mb:64 resemble-enhance <input_folder> <output_folder>" but still it ended in the same message " torch.cuda.OutOfMemoryError: CUDA out of memory. Tried to allocate 668.00 MiB. GPU 0 has a total capacty of 5.78 GiB of which 441.25 MiB is free. Including non-PyTorch memory, this process has 5.09 GiB memory in use. Of the allocated memory 4.96 GiB is allocated by PyTorch, and 11.45 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF"

I searched Issues before submitting to avoid duplicates, #37 is the only one related to cuda memory, the complain was the increase of memory usage but no mention that the process wasn't completed due to Vram full.

[CODE]System:
Kernel: 6.10.7-1-liquorix-amd64 [6.10-7~mx23ahs] arch: x86_64 bits: 64 compiler: gcc v: 12.2.0 parameters: audit=0
intel_pstate=disable
BOOT_IMAGE=/timeshift-btrfs/snapshots/2025-03-09_21-00-01/@/boot/vmlinuz-6.10.7-1-liquorix-amd64
root=UUID= ro rootflags=subvol=timeshift-btrfs/snapshots/2025-03-09_21-00-01/@ quiet
splash resume=UUID= resume_offset=4348782
Desktop: Xfce v: 4.20.0 tk: Gtk v: 3.24.38 info: xfce4-panel wm: xfwm v: 4.20.0 vt: 7
dm: LightDM v: 1.32.0 Distro: MX-23.5_ahs_x64 Libretto May 19 2024 base: Debian GNU/Linux 12
(bookworm)
Machine:
Type: Laptop System: Dell product: Precision 7740 v: N/A serial: Chassis:
type: 10 serial:
Mobo: Dell model: 0TX4KW v: A00 serial: UEFI: Dell v: 1.32.0
date: 04/01/2024
Battery:
ID-1: BAT0 charge: 68.1 Wh (100.0%) condition: 68.1/97.0 Wh (70.2%) volts: 12.3 min: 11.4
model: BYD DELL GW0K906 type: Li-poly serial: status: full
CPU:
Info: model: Intel Xeon E-2276M bits: 64 type: MT MCP arch: Coffee Lake level: v3 note: check
built: 2018 process: Intel 14nm family: 6 model-id: 0x9E (158) stepping: 0xD (13)
microcode: 0x100
Topology: cpus: 1x cores: 6 tpc: 2 threads: 12 smt: enabled cache: L1: 384 KiB
desc: d-6x32 KiB; i-6x32 KiB L2: 1.5 MiB desc: 6x256 KiB L3: 12 MiB desc: 1x12 MiB
Speed (MHz): avg: 1100 high: 1600 min/max: 800/2801 boost: enabled scaling:
driver: acpi-cpufreq governor: ondemand cores: 1: 1600 2: 1100 3: 1461 4: 1272 5: 800 6: 1600
7: 800 8: 800 9: 800 10: 800 11: 1368 12: 800 bogomips: 67198
Flags: avx avx2 ht lm nx pae sse sse2 sse3 sse4_1 sse4_2 ssse3 vmx[/CODE]