Skip to content

update mtia info in torchrec #3391

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

Closed
wants to merge 1 commit into from
Closed

update mtia info in torchrec #3391

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
6 changes: 3 additions & 3 deletions torchrec/distributed/planner/constants.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -77,9 +77,9 @@ def kernel_bw_lookup(
("cpu", EmbeddingComputeKernel.QUANT.value): 1 * ddr_mem_bw,
# TODO: Determine the correct value later. MTIA uses values same as CPU's.
# MTIA
("mtia", EmbeddingComputeKernel.DENSE.value): 0.5 * ddr_mem_bw,
("mtia", EmbeddingComputeKernel.FUSED.value): 1 * ddr_mem_bw,
("mtia", EmbeddingComputeKernel.QUANT.value): 1 * ddr_mem_bw,
("mtia", EmbeddingComputeKernel.DENSE.value): 0.5 * hbm_mem_bw,
("mtia", EmbeddingComputeKernel.FUSED.value): 1 * hbm_mem_bw,
("mtia", EmbeddingComputeKernel.QUANT.value): 1 * hbm_mem_bw,
# CUDA
("cuda", EmbeddingComputeKernel.DENSE.value): 0.5 * hbm_mem_bw,
("cuda", EmbeddingComputeKernel.FUSED.value): 1 * hbm_mem_bw,
Expand Down
5 changes: 4 additions & 1 deletion torchrec/distributed/planner/stats.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -705,7 +705,10 @@ def _log_rank_mem_usage_and_perf(
used_hbm_gb = bytes_to_gb(used_hbm[rank])
used_hbm_ratio = (
used_hbm[rank] / ((1 - reserved_hbm_percent) * device.storage.hbm)
if topology.compute_device == "cuda"
if (
topology.compute_device == "cuda"
or topology.compute_device == "mtia"
)
and ((1 - reserved_hbm_percent) * device.storage.hbm) != 0
else 0
)
Expand Down
Loading