Skip to content

fix(ci): bump NCCL to 2.28+ for TensorRT-LLM compatibility #777

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
slin1237 merged 1 commit into from
Mar 17, 2026
Merged

fix(ci): bump NCCL to 2.28+ for TensorRT-LLM compatibility #777

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 scripts/ci_install_trtllm.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5,7 +5,7 @@
# so we build from source (main branch) which compiles the C++
# extensions properly and includes the gRPC serve command.
#
# Cache version: 2 — rebuild to pick up gRPC stop_words fix (#11292)
# Cache version: 3 — rebuild for NCCL 2.28+ (required by TRT-LLM PR #12015)
#
# Prerequisites (expected on k8s-runner-gpu nodes):
# - NVIDIA driver 580+ (CUDA 13)
Expand Down Expand Up @@ -58,7 +58,7 @@ if [ -n "$CACHED_WHEEL" ] && [ -f "$CACHED_WHEEL" ]; then

# ── Install NCCL runtime ─────────────────────────────────────────────────
pip install --upgrade pip
pip install --no-cache-dir "nvidia-nccl-cu13>=2.27.7"
pip install --no-cache-dir "nvidia-nccl-cu13>=2.28.0"

# ── Install cached wheel ─────────────────────────────────────────────────
echo "Installing cached wheel..."
Expand Down Expand Up @@ -160,7 +160,7 @@ sudo ln -sf /usr/lib/x86_64-linux-gnu /usr/local/tensorrt/lib
# Use pip-installed NCCL 2.27+ which has both headers and libraries.
# This matches the working installation guide approach.
pip install --upgrade pip
pip install --no-cache-dir "nvidia-nccl-cu13>=2.27.7"
pip install --no-cache-dir "nvidia-nccl-cu13>=2.28.0"
Comment thread
slin1237 marked this conversation as resolved.

SITE_PACKAGES=$(python3 -c "import site; print(site.getsitepackages()[0])")

Expand Down
Loading