nvidia-smi -pm 1

kill -STOP 11402
nvidia-smi -r
kill -CONT 11402


        state_dict = torch.load(checkpoint_file, map_location=lambda storage, loc: storage)


!pip install flatbuffers==1


[tr][td=198]CUDA_?MANAGED_?FORCE_?DEVICE_?ALLOC[/td][td=82]0 or 1 (default is 0)[/td][td=212]Forces the driver to place all managed allocations in device memory.[/td][/tr]


[tr][td=198]CUDA_?DEVICE_?MAX_?CONNECTIONS[/td][td=82]1 to 32 (default is 8)[/td][td=212]Sets the number of compute and copy engine concurrent connections (work queues) from the host to each device of compute capability 3.5 and above.[/td][/tr]

[tr][td=198]CUDA_?CACHE_?PATH[/td][td=82]filepath[/td][td=212]Specifies the folder where the just-in-time compiler caches binary codes; the default values are:· on Windows, %APPDATA%\NVIDIA\ComputeCache,· on MacOS, $HOME/Library/Application\ Support/NVIDIA/ComputeCache,· on Linux, ~/.nv/ComputeCache[/td][/tr]

[tr][td=198]CUDA_?CACHE_?MAXSIZE[/td][td=82]integer (default is 33554432 (32 MB) and maximum is 4294967296 (4 GB))[/td][td=212]Specifies the size in bytes of the cache used by the just-in-time compiler. Binary codes whose size exceeds the cache size are not cached. Older binary codes are evicted from the cache to make room for newer binary codes if needed.[/td][/tr]