build: Bump vllm and deepgemm version (#2509)

container/Dockerfile.vllm

@@ -13,14 +13,14 @@ ARG RUNTIME_IMAGE="nvcr.io/nvidia/cuda"
 ARG RUNTIME_IMAGE_TAG="12.8.1-runtime-ubuntu24.04"
 
 # Make sure to update the dependency version in pyproject.toml when updating this
-ARG VLLM_REF="ba81acbdc1eec643ba815a76628ae3e4b2263b76"
+ARG VLLM_REF="77a6bf07aedf132aad2b6719f6d87abc5d3311ab"
 ARG TORCH_BACKEND="cu128"
 
 # Match 0.10.0 vLLM release
 # https://github.com/vllm-project/vllm/releases/tag/v0.10.0
 # Pinned to commit before https://github.com/deepseek-ai/DeepGEMM/pull/112 for DeepGEMM which seems to break on H100:
 # "RuntimeError: Failed: CUDA runtime error csrc/jit/kernel_runtime.hpp:108 '98'"
-ARG DEEPGEMM_REF="03d0be3"
+ARG DEEPGEMM_REF="f85ec64"
 ARG FLASHINF_REF="v0.2.8rc1"
 
 # Define general architecture ARGs for supporting both x86 and aarch64 builds.

container/deps/vllm/install_vllm.sh

@@ -20,12 +20,12 @@ set -euo pipefail
 
 # Parse arguments
 EDITABLE=true
-VLLM_REF="ba81acbdc1eec643ba815a76628ae3e4b2263b76"
+VLLM_REF="77a6bf07aedf132aad2b6719f6d87abc5d3311ab"
 VLLM_GIT_URL="https://github.com/vllm-project/vllm.git"
 MAX_JOBS=16
 INSTALLATION_DIR=/tmp
 ARCH=$(uname -m)
-DEEPGEMM_REF="03d0be3"
+DEEPGEMM_REF="f85ec64"
 FLASHINF_REF="v0.2.8rc1"
 TORCH_BACKEND="cu128"