chore: update nixl version to 0.4.1 (#2221)

Cargo.lock

@@ -4561,9 +4561,9 @@ dependencies = [
 
 [[package]]
 name = "nixl-sys"
-version = "0.4.0"
+version = "0.4.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "97f621270fd1ed8af5a8028a1945e6f7e612a38836ce82b720fe54222739df3c"
+checksum = "743ed1038b386b75451f9e0bba37cb2e3eea75873635268337d6531be99c9303"
 dependencies = [
  "bindgen 0.71.1",
  "cc",

container/Dockerfile.sglang

@@ -40,7 +40,7 @@ ARG ARCH
 ARG ARCH_ALT
 
 ARG NIXL_UCX_REF=v1.19.x
-ARG NIXL_REF=3c47a48955e6f96bd5d4fb43a9d80bb64722f8e4
+ARG NIXL_REF=0.4.1
 
 ENV NIXL_SRC_DIR=/opt/nixl
 ENV NIXL_PREFIX=/opt/nvidia/nvda_nixl

container/Dockerfile.sglang-wideep

@@ -71,7 +71,7 @@ RUN rm -rf /opt/hpcx/ucx && \
 
 ENV LD_LIBRARY_PATH=/usr/lib:/usr/local/ucx/lib:$LD_LIBRARY_PATH
 
-ARG NIXL_TAG=0.3.1
+ARG NIXL_TAG=0.4.1
 RUN git clone https://github.com/ai-dynamo/nixl.git && cd nixl && git checkout ${NIXL_TAG} && pip install --break-system-packages . --config-settings=setup-args="-Ducx_path=/usr/local/ucx"
 
 WORKDIR /sgl-workspace

container/Dockerfile.tensorrt_llm

@@ -45,7 +45,7 @@ ARG ARCH
 ARG ARCH_ALT
 
 ARG NIXL_UCX_REF=v1.19.x
-ARG NIXL_REF=3c47a48955e6f96bd5d4fb43a9d80bb64722f8e4
+ARG NIXL_REF=0.4.1
 
 ENV NIXL_SRC_DIR=/opt/nixl
 ENV NIXL_PREFIX=/opt/nvidia/nvda_nixl

container/Dockerfile.vllm

@@ -81,7 +81,7 @@ RUN apt-get update -y && \
     rm -rf /var/lib/apt/lists/*
 
 ARG NIXL_UCX_REF=v1.19.x
-ARG NIXL_REF=3c47a48955e6f96bd5d4fb43a9d80bb64722f8e4
+ARG NIXL_REF=0.4.1
 
 ENV NIXL_SRC_DIR=/opt/nixl
 ENV NIXL_PREFIX=/opt/nvidia/nvda_nixl

container/build.sh

@@ -113,7 +113,7 @@ NONE_BASE_IMAGE_TAG="24.04"
 SGLANG_BASE_IMAGE="nvcr.io/nvidia/cuda-dl-base"
 SGLANG_BASE_IMAGE_TAG="25.01-cuda12.8-devel-ubuntu24.04"
 
-NIXL_REF=3c47a48955e6f96bd5d4fb43a9d80bb64722f8e4
+NIXL_REF=0.4.1
 NIXL_UCX_EFA_REF=7ec95b95e524a87e81cac92f5ca8523e3966b16b
 
 NO_CACHE=""
@@ -389,8 +389,6 @@ ARCH="amd64"
 if [[ "$PLATFORM" == *"linux/arm64"* ]]; then
     ARCH="arm64"
     BUILD_ARGS+=" --build-arg ARCH=arm64 --build-arg ARCH_ALT=aarch64 "
-    # TEMP: Pin to nixl 0.3.1 for arm build, since 0.4.0 fails
-    NIXL_REF=3503658e71143b56f9d5b1b440d84a94b9c41af8
 fi
 
 # Update DOCKERFILE if framework is VLLM

lib/bindings/python/Cargo.lock

@@ -2912,9 +2912,9 @@ dependencies = [
 
 [[package]]
 name = "nixl-sys"
-version = "0.4.0"
+version = "0.4.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "97f621270fd1ed8af5a8028a1945e6f7e612a38836ce82b720fe54222739df3c"
+checksum = "743ed1038b386b75451f9e0bba37cb2e3eea75873635268337d6531be99c9303"
 dependencies = [
  "bindgen",
  "cc",

lib/llm/Cargo.toml

@@ -90,7 +90,7 @@ rayon = "1"
 dialoguer = { version = "0.11", default-features = false, features = ["editor", "history"] }
 
 # block_manager
-nixl-sys = {version = "0.4.0", optional = true }
+nixl-sys = {version = "0.4.1", optional = true }
 cudarc = { version = "0.16.2", features = ["cuda-12020"], optional = true }
 ndarray = { version = "0.16", optional = true }
 nix = { version = "0.26", optional = true }

pyproject.toml

@@ -66,13 +66,13 @@ trtllm =[
 
 vllm = [
     "uvloop",
-    "nixl",
+    "nixl<=0.4.1",
     "vllm==0.10.0",
 ]
 
 sglang = [
     "uvloop",
-    "nixl",
+    "nixl<=0.4.1",
     "sglang[all]==0.4.9.post6",
 ]