Dynamo Tensorrt-LLM Runtime

NVIDIA

Container

NVIDIA

Dynamo Tensorrt-LLM Runtime

The Dynamo TensorRT-LLM runtime image is a containerized build of Dynamo + TensorRT-LLM which serves as the base runtime environment for tensorrt-llm based inference with Dynamo's distributed inference framework.

Layer	Label		Created
sha256:8db50556d0fb0ccb425c9b3efa2d90de9e6407e46fdf33706493dda7d1bc466c	COPY	`/legal /legal`	10/21/2026 8:33 PM UTC
sha256:a3ed95caeb02ffe68cdd9fd84406680ae93d633cb16422d00e8a7c22955b46d4	ENV	`NVIDIA_DRIVER_CAPABILITIES=video,compute,utility`	10/21/2026 8:32 PM UTC
sha256:a3ed95caeb02ffe68cdd9fd84406680ae93d633cb16422d00e8a7c22955b46d4	CMD	`/bin/bash`	10/21/2026 8:32 PM UTC
sha256:a3ed95caeb02ffe68cdd9fd84406680ae93d633cb16422d00e8a7c22955b46d4	ENTRYPOINT	`—`	10/21/2026 8:32 PM UTC
sha256:a3ed95caeb02ffe68cdd9fd84406680ae93d633cb16422d00e8a7c22955b46d4	USER	`dynamo`	10/21/2026 8:32 PM UTC
sha256:a3ed95caeb02ffe68cdd9fd84406680ae93d633cb16422d00e8a7c22955b46d4	ENV	`DYNAMO_COMMIT_SHA=f65e8fdf438e92806a1e586a8cd6ba9100a0cdfb`	10/21/2026 8:32 PM UTC
sha256:a3ed95caeb02ffe68cdd9fd84406680ae93d633cb16422d00e8a7c22955b46d4	ARG	`DYNAMO_COMMIT_SHA=f65e8fdf438e92806a1e586a8cd6ba9100a0cdfb`	10/21/2026 8:32 PM UTC
sha256:4f4fb700ef54461cfa02571ae0db9a0dc1e0cdb5577484a6d75e68dc38e8acc1	WORKDIR	`/workspace`	10/21/2026 8:32 PM UTC
sha256:a3ed95caeb02ffe68cdd9fd84406680ae93d633cb16422d00e8a7c22955b46d4	ENV	DYNAMO_HOME=/workspace HOME=/home/dynamo VIRTUAL_ENV=/opt/dynamo/venv PATH=/opt/dynamo/venv/bin:/opt/dynamo/mpi/bin:/opt/uv/bin:/usr/local/bin/etcd:/opt/trtllm/.venv-3.12/bin:/usr/local/lib/python3.12/dist-packages/torch_tensorrt/bin:/usr/local/nvidia/bin:/usr/local/cuda/bin:/usr/local/mpi/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/local/ucx/bin:/opt/amazon/efa/bin:/opt/tensorrt/bin IMAGEIO_FFMPEG_EXE=/usr/local/bin/ffmpeg LD_PRELOAD=/opt/dynamo/libstdc++.so.6:/opt/trtllm/tensorrt_llm/libs/nixl/libnixl.so LD_LIBRARY_PATH=/opt/dynamo/mpi/lib:/usr/local/tensorrt/lib:/usr/local/lib/python3.12/dist-packages/torch/lib:/usr/local/lib/python3.12/dist-packages/torch_tensorrt/lib:/usr/local/cuda/compat/lib:/usr/local/nvidia/lib:/usr/local/nvidia/lib64 OPAL_PREFIX=/opt/dynamo/mpi NIXL_PLUGIN_DIR=/opt/trtllm/tensorrt_llm/libs/nixl/plugins NIXL_VERSION=	10/21/2026 8:32 PM UTC
sha256:4f4fb700ef54461cfa02571ae0db9a0dc1e0cdb5577484a6d75e68dc38e8acc1	RUN	`set -eu; tb=$(/usr/bin/python3 -c 'import os, triton.backends.nvidia as b; print(os.path.dirname(b.__file__))'); if [ ! -f "$tb/include/cuda.h" ]; then echo "ERROR: Triton CUDA header missing after runtime overlay: $tb/include/cuda.h" >&2; exit 1; fi; for tool in ptxas cuobjdump nvdisasm; do if [ ! -x "$tb/bin/$tool" ]; then echo "ERROR: Triton CUDA tool missing or not executable after runtime overlay: $tb/bin/$tool" >&2; exit 1; fi; done`	10/21/2026 8:32 PM UTC