viable/strict/1791914049: [CUDA] Migrate UVM MemPool to Python Allocator (#199639)
- PyTorch: 2266 events in the last 90 days
- PyTorch: 2236th Release in the last 90 days
- Previous: earlier the same day · viable/strict/1791910592
What happened
Replace existing ctypes based UVM mempool to use the new Python allocator introduced in #199638 and #200427 __ Ran existing tests locally: • python test/test_cuda.py -k use_uvm -v test_use_uvm ... ok test_use_uvm_advice_error_frees_allocation ... ok test_use_uvm_backward ... ok test_use_uvm_numerics ... ok test_use_uvm_oom_uses_caching_allocator_recovery ... ok test_use_uvm_tensor_outlives_context ... ok -----------…
Summary assembled by rule from the sources below