RuntimeError: CUDA out of memory with a huge amount of free memory

While training the model for image colorization, I encountered the following problem:

RuntimeError: CUDA out of memory. Tried to allocate 304.00 MiB (GPU 0; 8.00 GiB total capacity; 142.76 MiB already allocated; 6.32 GiB free; 158.00 MiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF

As we can see, the error occurs when trying to allocate 304 MiB of memory, while 6.32 GiB is free! What is the problem? As I can see, the suggested option is to set max_split_size_mb to avoid fragmentation. Will it help and how to do it correctly?
My batch size = 40
This is my version of PyTorch:

torch==1.10.2+cu113

torchvision==0.11.3+cu113

torchaudio===0.10.2+cu113

Could you post a minimal, executable code snippet to reproduce this issue as well as the output of python -m torch.utils.collect_env, please?

Thank you for the quick response and I hope for your help, it is important for my thesis at the university.
This is python -m torch.utils.collect_env console output:

PyTorch version: 1.10.2+cu113
Is debug build: False
CUDA used to build PyTorch: 11.3
ROCM used to build PyTorch: N/A

OS: Microsoft Windows 10 Pro
GCC version: Could not collect
Clang version: Could not collect
CMake version: Could not collect
Libc version: N/A

Python version: 3.6.13 |Anaconda, Inc.| (default, Mar 16 2021, 11:37:27) [MSC v.1916 64 bit (AMD64)] (64-bit runtime)
Python platform: Windows-10-10.0.19041-SP0
Is CUDA available: True
CUDA runtime version: 11.3.109
GPU models and configuration: GPU 0: NVIDIA GeForce RTX 2070
Nvidia driver version: 511.23
cuDNN version: Could not collect
HIP runtime version: N/A
MIOpen runtime version: N/A

Versions of relevant libraries:
[pip3] numpy==1.19.2
[pip3] torch==1.10.2+cu113
[pip3] torchaudio==0.10.2+cu113
[pip3] torchvision==0.11.3+cu113
[conda] blas 1.0 mkl
[conda] cpuonly 1.0 0 pytorch
[conda] cudatoolkit 11.3.1 h59b6b97_2
[conda] mkl 2020.2 256
[conda] mkl-service 2.3.0 py36h196d8e1_0
[conda] mkl_fft 1.3.0 py36h46781fe_0
[conda] mkl_random 1.1.1 py36h47e9c7a_0
[conda] numpy 1.19.2 py36hadc3359_0
[conda] numpy-base 1.19.2 py36ha3acd2a_0
[conda] pytorch-mutex 1.0 cpu pytorch
[conda] torch 1.10.2+cu113 pypi_0 pypi
[conda] torchaudio 0.10.2+cu113 pypi_0 pypi
[conda] torchvision 0.11.3 pypi_0 pypi

I don’t really know if I can post GitHub links here, but that would be the best explanation for your request. I’m trying to run code from this repository, specifically this script:
GitHub repository

Thanks for the update. Could you try to narrow down the code and post a minimal, executable code snippet, which we could use to reproduce the issue?