PyTorch Community, The Meta team are happy to make our 2025 H1 roadmaps available. We plan on a half year basis and globally optimize across the things we do for our users here at Meta and across the wider community. We’ve removed a few references to internal systems and teams but are eager to shar…

Meta PyTorch Team 2025 H1 Roadmaps

johnnynunez February 20, 2025, 5:19pm 2

Will you imrpove for upcoming devices with cuda with unified memory? Also for jetson orin/thor, digits, thor etc…

cudaMallocManaged() in all places…

Topic		Replies	Views
Understanding the difference between the caching behavior of cuda caching allocator and pluggable allocator	7	559	August 2, 2025
Meta PyTorch Team 2025 H2 Roadmaps	6	2800	August 28, 2025
Meta PyTorch Team 2024 H2 Roadmaps	20	23399	February 19, 2025
FSDP & CUDACachingAllocator: an outsider newb perspective distributed	10	8484	December 13, 2024
Memory operations on a custom backend hardware-backends	4	1236	July 5, 2022