Can the Orin AGX 64GB model use version 0.20 of vllm, or is version 0.17 or higher also acceptable?

[WGY2e]

Search before asking

• I have searched the jetson-containers issues and found no similar feature requests.

Question

My current VLLM is version 0.16 and only has the standard OpenAI interface. I want to upgrade to version 0.20 of VLLM which can support the Claude API. I would like to know how to complete the compilation of the image on the Orin AGX.

Additional

No response

[project2go]

looks like 0.20 added to vllm config.py but... build vllm failed with cutlass dependency?

testing cutlass...
Traceback (most recent call last):
File "/test/test_runtime.py", line 3, in
import cutlass
ModuleNotFoundError: No module named 'cutlass'
[23:19:51] =====================================================================================
[23:19:51] =====================================================================================
[23:19:51] 💣 jetson-containers build failed after 145.7 seconds (2.4 minutes)
[23:19:51] Error: Command 'docker run -t --rm --network=host --privileged --shm-size=8g --gpus=all --env NVIDIA_DRIVER_CAPABILITIES=all --volume /home/dev/jetson-containers/packages/cuda/cutlass:/test --volume /home/dev/jetson-containers/data:/data vllm:r36.5.tegra-aarch64-cu126-22.04-cutlass /bin/bash -c 'python3 /test/test_runtime.py' 2>&1 | tee /home/dev/jetson-containers/logs/20260405_231725/test/19-1_vllm_r36.5.tegra-aarch64-cu126-22.04-cutlass_test_runtime.py.txt; exit ${PIPESTATUS[0]}' returned non-zero exit status 1.
[23:19:51] =====================================================================================
[23:19:51] =====================================================================================
[23:19:51] Failed building: vllm

Traceback (most recent call last):
File "/home/dev/jetson-containers/jetson_containers/build.py", line 136, in
build_container(**vars(args))
File "/home/dev/jetson-containers/jetson_containers/container.py", line 280, in build_container
test_container(container_name, pkg, simulate, build_idx=idx)
File "/home/dev/jetson-containers/jetson_containers/container.py", line 487, in test_container
status = subprocess.run(cmd.replace(NEWLINE, ' '), executable='/bin/bash', shell=True, check=True)
File "/usr/lib/python3.10/subprocess.py", line 526, in run
raise CalledProcessError(retcode, process.args,
subprocess.CalledProcessError: Command 'docker run -t --rm --network=host --privileged --shm-size=8g --gpus=all --env NVIDIA_DRIVER_CAPABILITIES=all --volume /home/dev/jetson-containers/packages/cuda/cutlass:/test --volume /home/dev/jetson-containers/data:/data vllm:r36.5.tegra-aarch64-cu126-22.04-cutlass /bin/bash -c 'python3 /test/test_runtime.py' 2>&1 | tee /home/dev/jetson-containers/logs/20260405_231725/test/19-1_vllm_r36.5.tegra-aarch64-cu126-22.04-cutlass_test_runtime.py.txt; exit ${PIPESTATUS[0]}' returned non-zero exit status 1.

[WGY2e]

looks like 0.20 added to vllm config.py but... build vllm failed with cutlass dependency?

testing cutlass... Traceback (most recent call last): File "/test/test_runtime.py", line 3, in import cutlass ModuleNotFoundError: No module named 'cutlass' [23:19:51] ===================================================================================== [23:19:51] ===================================================================================== [23:19:51] 💣 failed after 145.7 seconds (2.4 minutes) [23:19:51] Error: Command 'docker run -t --rm --network=host --privileged --shm-size=8g --gpus=all --env NVIDIA_DRIVER_CAPABILITIES=all --volume /home/dev/jetson-containers/packages/cuda/cutlass:/test --volume /home/dev/jetson-containers/data:/data vllm:r36.5.tegra-aarch64-cu126-22.04-cutlass /bin/bash -c 'python3 /test/test_runtime.py' 2>&1 | tee /home/dev/jetson-containers/logs/20260405_231725/test/19-1_vllm_r36.5.tegra-aarch64-cu126-22.04-cutlass_test_runtime.py.txt; exit ${PIPESTATUS[0]}' returned non-zero exit status 1. [23:19:51] ===================================================================================== [23:19:51] ===================================================================================== [23:19:51] Failed building: vllmjetson-containers build

Traceback (most recent call last): File "/home/dev/jetson-containers/jetson_containers/build.py", line 136, in build_container(**vars(args)) File "/home/dev/jetson-containers/jetson_containers/container.py", line 280, in build_container test_container(container_name, pkg, simulate, build_idx=idx) File "/home/dev/jetson-containers/jetson_containers/container.py", line 487, in test_container status = subprocess.run(cmd.replace(NEWLINE, ' '), executable='/bin/bash', shell=True, check=True) File "/usr/lib/python3.10/subprocess.py", line 526, in run raise CalledProcessError(retcode, process.args, subprocess.CalledProcessError: Command 'docker run -t --rm --network=host --privileged --shm-size=8g --gpus=all --env NVIDIA_DRIVER_CAPABILITIES=all --volume /home/dev/jetson-containers/packages/cuda/cutlass:/test --volume /home/dev/jetson-containers/data:/data vllm:r36.5.tegra-aarch64-cu126-22.04-cutlass /bin/bash -c 'python3 /test/test_runtime.py' 2>&1 | tee /home/dev/jetson-containers/logs/20260405_231725/test/19-1_vllm_r36.5.tegra-aarch64-cu126-22.04-cutlass_test_runtime.py.txt; exit ${PIPESTATUS[0]}' returned non-zero exit status 1.

[WGY2e]

我已经构建完成了，内部缺少非常多的依赖，很多的库已经不能用了，可能需要自己去找