Skip to content

[None][fix] change logging for weight loading on unified memory #9177

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

Merged
farazkh80 merged 4 commits into from
Nov 19, 2025
+3 −3
Merged

[None][fix] change logging for weight loading on unified memory #9177

Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
1f15137
Add cpath for mpi env var fix
farazkh80 Nov 14, 2025
a4a3484
[None][fix] Reduce load_weight_shard warning message on integrated sy…
SimengLiu-nv Nov 18, 2025
b65230a
undo unnecessary change
farazkh80 Nov 18, 2025
d3b2213
undo dockerfile expose
farazkh80 Nov 19, 2025
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 tensorrt_llm/_torch/modules/linear.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -75,9 +75,9 @@ def load_weight_shard(
# For integrated GPU systems (e.g., DGX Spark), CPU and GPU share limited physical memory.
# Avoiding device transfers reduces memory consumption and unnecessary data copies,
# enabling support for larger models on memory-constrained systems.
logger.warning(
f"[load_weight_shard] Skipping device transfer from {weight.device} to {device} on integrated GPU to conserve shared memory."
)
logger.warning_once(
f"[load_weight_shard] Skipping device transfer from {weight.device} to {device} on integrated GPU to conserve shared memory.",
key="load_weight_shard_skip_device_transfer_with_integrated_gpu")
device = weight.device
if isinstance(weight, torch.Tensor):
tensor_shape = weight.shape
Expand Down