Package Details: python-vllm-rocm 0.24.0-1

Git Clone URL: https://aur.archlinux.org/python-vllm-rocm.git (read-only, click to copy)
Package Base: python-vllm-rocm
Description: high-throughput and memory-efficient inference and serving engine for LLMs (ROCm support)
Upstream URL: https://github.com/vllm-project/vllm
Licenses: Apache-2.0
Provides: python-vllm
Submitter: davispuh
Maintainer: davispuh
Last Packager: davispuh
Votes: 3
Popularity: 0.26
First Submitted: 2026-02-24 22:16 (UTC)
Last Updated: 2026-07-02 19:28 (UTC)

Latest Comments

1 2 Next › Last »

chiz commented on 2026-07-12 06:17 (UTC) (edited on 2026-07-12 06:19 (UTC) by chiz)

python_aotriton is 0.13b-1

Building for ROCM=gfx1151
Traceback (most recent call last):
  File "/home/chi/.cache/yay/python-vllm-rocm/src/vllm/setup.py", line 17, in <module>
    import torch
  File "/usr/lib/python3.14/site-packages/torch/__init__.py", line 444, in <module>
    from torch._C import *  # noqa: F403
    ^^^^^^^^^^^^^^^^^^^^^^
ImportError: libaotriton_v2.so.0.11.2: cannot open shared object file: No such file or directory

leuko commented on 2026-07-05 17:59 (UTC) (edited on 2026-07-05 18:20 (UTC) by leuko)

In 0.24.0-1, I got errors during build. I had to do the following to fix it:

  1. Install :

python-setuptools-rust python-openai-harmony python-model-hosting-container-standards python-jmespath python-compressed-tensors # is orphaned. Had to comment test section and update version number

  1. Add the following in build right after the cd command. 
# In case ROCM was installed in the same session for the first time, then `rocm*`
# commands will not be available and (1) `setup.py` will fail (2) The local GPUs
# will not be detected, so binaries for all GPU architectures will be built. So
# include `rocm.sh` manually:
. /etc/profile

davispuh commented on 2026-05-23 21:03 (UTC)

That's not vLLM issue but python-transformers. You need to patch it, follow comment instructions => https://aur.archlinux.org/packages/python-transformers#comment-1069558

Slug commented on 2026-05-23 01:14 (UTC) (edited on 2026-05-23 01:18 (UTC) by Slug)

hello! Thank you for maintaining this package

python-transformers fails during check() phase due to a strict version constraint on tokenizers, blocking the build of python-vllm-rocm.

ImportError: tokenizers>=0.22.0,<=0.23.0 is required for a normal functioning of this module,but found tokenizers==0.23.1.

check() fails immediately and build aborts, preventing the installation of python-vllm-rocm

Attempted workaround

  1. AUR build bypass

yay -S python-transformers --mflags "--nocheck" yay -S python-vllm-rocm --mflags "--nocheck"

Result:Build succeeds. Runtime execution attempt :

python -m vllm.entrypoints.openai.api_server

Result:immediate ImportError from transformers ; execution aborts before any model loading

GPU: AMD Radeon RX 7900 XTX (gfx1100) ROCm: 7.2.3

davispuh commented on 2026-04-30 00:10 (UTC)

@Orion-zhen I don't think that's good idea. All ROCM packages will have such issue and it's different for every shell. That would only work for bash. It's better for your build script to do that.

@chiz that happens because intel-oneapi-mkl was updated so you also need to update python-pytorch-opt-rocm (or rebuild your PyTorch).

chiz commented on 2026-04-26 16:38 (UTC)

there is a error:

OSError: libmkl_intel_lp64.so.2: cannot open shared object file: No such file or directory

Orion-zhen commented on 2026-04-06 13:21 (UTC) (edited on 2026-04-06 13:24 (UTC) by Orion-zhen)

Please add this in build() in case that a newly installed rocm can't be recognized:

  if [[ -z "${ROCM_PATH}" ]]; then
    source /etc/profile
  fi

I failed to build this package in GitHub Action runner.

davispuh commented on 2026-04-05 15:41 (UTC)

python-cbor2 is indeed required, added it as dependency.

But others I don't have them installed and Qwen3.5-4B works fine without any issues. I guess it depends on what model you want to run.

Also I tested that python-compressed-tensors builds and installs with --nocheck (only check step fails for me). It's needed for gpt-oss-20b but I couldn't get it to run on my RX 7900 XTX (No MXFP4 MoE backend supports the deployment configuration).

bnjbvr commented on 2026-03-30 20:47 (UTC) (edited on 2026-03-30 20:48 (UTC) by bnjbvr)

Fwiw, I had to also install the following packages, so as to try to make it work:

  • python-cbor2
  • python-openai-harmony
  • python-model-hosting-container-standards
  • python-jmespath
  • python-compressed-tensors

Otherwise, the command vllm serve … wouldn't work. That being said, python-compressed-tensors didn't seem to install successfully, and the package is marked as orphaned, so I eventually had to drop the whole thing, unfortunately.

1 2 Next › Last »