Update dynamo-xla to the new torch-xla version and add DLPack support#25

Merged

yitongh merged 22 commits intoaccfrom dev/dynamo_dlpack

Dec 11, 2024

+2,472-696

Commits on Dec 11, 2024

Add dlpack support (pytorch#7025 )

vanbasten23
authored and
yitongh
committed
Make from_dlpack handle cuda synchronization implicitly for input tensors that have __dlpack__ and __dlpack_device__ attributes. (pytorch#7125 )

vanbasten23
authored and
yitongh
committed
Reuse DLDeviceType. (pytorch#7163 )

vanbasten23
authored and
yitongh
committed
Automatically move CUDA non XLA Tensors to XLA Device and back to CUDA device (pytorch#6644 )

changm
authored and
yitongh
committed
Add a DynamoSyncInputExecuteTime counter (pytorch#6813 )

JackCaoG
authored and
yitongh
committed
Fix runtime error when run dynamo with a profiler scope (pytorch#6913 )

JackCaoG
authored and
yitongh
committed
Add function for retrieving fallback operations. (pytorch#7116 )

ysiraichi
authored and
yitongh
committed
Integrate dlpack to dynamo. (pytorch#7173 )

vanbasten23
authored and
yitongh
committed
Add support for dynamic shape in dynamo (pytorch#7676 )

authored and
yitongh
committed
Optimize dynamo dynamic shape caching (pytorch#7726 )

JackCaoG
authored and
yitongh
committed
Fix the crash when symint is part of the output with dynamic torch.compile (pytorch#7735 )

JackCaoG
authored and
yitongh
committed
reenable dynamo dynamic shape test (pytorch#7775 )

JackCaoG
authored and
yitongh
committed
Support mark_dynamic (pytorch#7812 )

JackCaoG
authored and
yitongh
committed
DYNAMO RNG seed update optimization (pytorch#7884 )

JackCaoG
authored and
yitongh
committed
CPU time optimization for GraphInputMatcher (pytorch#7895 )

JackCaoG
authored and
yitongh
committed
Optimize _split_xla_args_tensor_sym_constant (pytorch#7900 )

JackCaoG
authored and
yitongh
committed
Fix the crash with copy op in dynamo (pytorch#7902 )

JackCaoG
authored and
yitongh
committed
Move the dynamo bridge to the _dynamo repo (pytorch#7909 )

JackCaoG
authored and
yitongh
committed
add dynamo config skip_input_data_check (pytorch#7913 )

JackCaoG
authored and
yitongh
committed
In dynamo optim_mode avoid unnecessary set_attr (pytorch#7915 )

JackCaoG
authored and
yitongh
committed
fix dynamo inplace copy (pytorch#7933 )

zpcore
authored and
yitongh
committed
fix bug for compile related to dlpack
yitongh
committed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Update dynamo-xla to the new torch-xla version and add DLPack support#25

Commits on Dec 11, 2024

Add dlpack support (pytorch#7025 )

Make from_dlpack handle cuda synchronization implicitly for input tensors that have dlpack and __dlpack_device__ attributes. (pytorch#7125 )

Reuse DLDeviceType. (pytorch#7163 )

Automatically move CUDA non XLA Tensors to XLA Device and back to CUDA device (pytorch#6644 )

Add a `DynamoSyncInputExecuteTime` counter (pytorch#6813 )

Fix runtime error when run dynamo with a profiler scope (pytorch#6913 )

Add function for retrieving fallback operations. (pytorch#7116 )

Integrate dlpack to dynamo. (pytorch#7173 )

Add support for dynamic shape in dynamo (pytorch#7676 )

Optimize dynamo dynamic shape caching (pytorch#7726 )

Fix the crash when symint is part of the output with dynamic torch.compile (pytorch#7735 )

reenable dynamo dynamic shape test (pytorch#7775 )

Support mark_dynamic (pytorch#7812 )

DYNAMO RNG seed update optimization (pytorch#7884 )

CPU time optimization for GraphInputMatcher (pytorch#7895 )

Optimize _split_xla_args_tensor_sym_constant (pytorch#7900 )

Fix the crash with copy op in dynamo (pytorch#7902 )

Move the dynamo bridge to the _dynamo repo (pytorch#7909 )

add dynamo config skip_input_data_check (pytorch#7913 )

In dynamo optim_mode avoid unnecessary set_attr (pytorch#7915 )

fix dynamo inplace copy (pytorch#7933 )

fix bug for compile related to dlpack