Skip to content

[release/2.5] AMDSMI/layernorm cherry picks #1764

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 2 commits into
base: release/2.5
Choose a base branch
from
Open

[release/2.5] AMDSMI/layernorm cherry picks #1764

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
6024e85
[ROCm] Correct numerical issues in layer norm backwards kernel (#140259)
jataylo Nov 11, 2024
aea2827
[ROCm] AMDSMI memory usage unification (#139900)
jataylo Nov 21, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 2 additions & 2 deletions aten/src/ATen/native/cuda/layer_norm_kernel.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -840,8 +840,8 @@ void cuLoadWriteStridedInputs(
{
int i1 = i1_block+thr_load_row_off;
if (i1 < i1_end) {
T curr_mean = mean[i1];
T curr_rstd = rstd[i1];
T_ACC curr_mean = mean[i1];
T_ACC curr_rstd = rstd[i1];
for (int k = 0; k < blockDim.y; ++k) {
int i2 = i2_off + k;
int load_idx = i1*N+i2;
Expand Down
3 changes: 2 additions & 1 deletion torch/cuda/__init__.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1092,7 +1092,8 @@ def _get_amdsmi_device_index(device: Optional[Union[int, Device]]) -> int:
def _get_amdsmi_memory_usage(device: Optional[Union[Device, int]] = None) -> int:
handle = _get_amdsmi_handler()
device = _get_amdsmi_device_index(device)
return amdsmi.amdsmi_get_gpu_vram_usage(handle)["vram_used"]
handle = amdsmi.amdsmi_get_processor_handles()[device]
return amdsmi.amdsmi_get_gpu_activity(handle)["umc_activity"]


def _get_amdsmi_utilization(device: Optional[Union[Device, int]] = None) -> int:
Expand Down