Skip to content

feat(mindtorch_v2): add CUDA backend MVP with explicit dispatch keys#2753

Merged
lvyufeng merged 2 commits into
candle-org:masterfrom
lvyufeng:mtv2-cuda-backend-mvp
Mar 8, 2026
Merged

feat(mindtorch_v2): add CUDA backend MVP with explicit dispatch keys#2753
lvyufeng merged 2 commits into
candle-org:masterfrom
lvyufeng:mtv2-cuda-backend-mvp

Conversation

@lvyufeng

@lvyufeng lvyufeng commented Mar 8, 2026

Copy link
Copy Markdown
Collaborator

Summary

  • add a pure-Python mindtorch_v2 CUDA backend MVP covering runtime, storage, transfer, and creation APIs
  • upgrade CUDA dispatch from PrivateUse1 to explicit DispatchKey.CUDA and DispatchKey.AutogradCUDA
  • add focused CUDA backend and dispatch regression coverage plus update the implementation/design docs

Test Plan

  • PYTHONPATH=src pytest -q tests/mindtorch_v2/contract/test_schema_registration_order.py
  • PYTHONPATH=src pytest -q tests/mindtorch_v2/contract/test_schema_coverage.py
  • PYTHONPATH=src pytest -q tests/mindtorch_v2/test_dispatch_keys.py tests/mindtorch_v2/test_dispatch_registry.py tests/mindtorch_v2/test_dispatch_resolution.py tests/mindtorch_v2/test_device_transfer.py tests/mindtorch_v2/test_cuda_backend.py tests/mindtorch_v2/test_creation.py tests/mindtorch_v2/test_dtype_device.py

@lvyufeng
lvyufeng force-pushed the mtv2-cuda-backend-mvp branch from 2853885 to 5e4e6e4 Compare March 8, 2026 06:21
@lvyufeng
lvyufeng merged commit 084e891 into candle-org:master Mar 8, 2026
4 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant