torch-mlir/python/torch_mlir
Dixin Zhou 8f52f5a4ed
[Fx Importer] fix mutation importer with non persistent buffer (#3798)
A non-persistent buffer will not be a part of this module’s
`state_dict`. Hence when setting `experimental_support_mutation=True`
and have non-persistent buffer, the current fx importer will fail to
retrieve a value from `state_dict` and produce `torch.constant.none` to
represent the buffer. This fix get value of non-persistent buffer from
the module's `constants`.

---------

Co-authored-by: Dixin Zhou <dzhou@vdi-ahddp-020.dhcp.mathworks.com>
2024-11-01 02:20:32 +08:00
..
_mlir_libs [ci] Upgrade to new runners and disable unsupported jobs. (#2818) 2024-01-27 18:35:45 -08:00
dialects Upstream the ONNX importer. (#2636) 2023-12-12 19:02:51 -08:00
extras [Fx Importer] fix mutation importer with non persistent buffer (#3798) 2024-11-01 02:20:32 +08:00
tools/import_onnx Bump Onnx Version to 1.16.1 (#3515) 2024-07-01 22:15:45 +05:30
compiler_utils.py [Torch] Extract TensorPlaceholder to a common interface (#3668) 2024-08-27 23:31:28 +08:00
fx.py [fx] Fix importing and tests for quantized conv (#3809) 2024-10-22 18:37:57 +02:00