implementation detail · not yet filed
PyTorch expandable segments allocator (PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True) for loading checkpoints
Also called expandable allocator.
- source
- 1
- model
- 1
- lab adopt it
- 1
- strongest
- used
How sources treat it
One count per evidence span, weakest treatment to strongest.
used 1
Documented in
Evidence
1 span quoted from the sources, strongest treatment first.
make sure to turn on the expandable allocator to avoid crashes when loading weights from the checkpoint.
usedsoftware implementationin gpt-oss reference implementationsOpenAI