Trying to run the tests:
python perf-kernels/flash-attention.py Traceback (most recent call last): File "/home/rohit/vllm/triton/python/perf-kernels/flash-attention.py", line 28, in <module> import triton File "/home/rohit/miniconda3/envs/vllm/lib/python3.10/site-packages/triton-2.1.0-py3.10-linux-x86_64.egg/triton/__init__.py", line 8, in <module> from .runtime import ( File "/home/rohit/miniconda3/envs/vllm/lib/python3.10/site-packages/triton-2.1.0-py3.10-linux-x86_64.egg/triton/runtime/__init__.py", line 1, in <module> from .autotuner import (Autotuner, Config, Heuristics, OutOfResources, autotune, heuristics) File "/home/rohit/miniconda3/envs/vllm/lib/python3.10/site-packages/triton-2.1.0-py3.10-linux-x86_64.egg/triton/runtime/autotuner.py", line 7, in <module> from ..testing import do_bench, do_bench_cudagraph File "/home/rohit/miniconda3/envs/vllm/lib/python3.10/site-packages/triton-2.1.0-py3.10-linux-x86_64.egg/triton/testing.py", line 8, in <module> from . import language as tl File "/home/rohit/miniconda3/envs/vllm/lib/python3.10/site-packages/triton-2.1.0-py3.10-linux-x86_64.egg/triton/language/__init__.py", line 4, in <module> from . import math File "/home/rohit/miniconda3/envs/vllm/lib/python3.10/site-packages/triton-2.1.0-py3.10-linux-x86_64.egg/triton/language/math.py", line 5, in <module> from . import core File "/home/rohit/miniconda3/envs/vllm/lib/python3.10/site-packages/triton-2.1.0-py3.10-linux-x86_64.egg/triton/language/core.py", line 8, in <module> from .._C.libtriton.triton import ir ImportError: /home/rohit/miniconda3/envs/vllm/bin/../lib/libstdc++.so.6: versionGLIBCXX_3.4.30' not found (required by /home/rohit/miniconda3/envs/vllm/lib/python3.10/site-packages/triton-2.1.0-py3.10-linux-x86_64.egg/triton/_C/libtriton.so)
`
help requested to resolve this.
I installed AMD-Triton using this guide: https://rocm.docs.amd.com/en/latest/how-to/llm-fine-tuning-optimization/model-acceleration-libraries.html
Trying to run the tests:
python perf-kernels/flash-attention.py Traceback (most recent call last): File "/home/rohit/vllm/triton/python/perf-kernels/flash-attention.py", line 28, in <module> import triton File "/home/rohit/miniconda3/envs/vllm/lib/python3.10/site-packages/triton-2.1.0-py3.10-linux-x86_64.egg/triton/__init__.py", line 8, in <module> from .runtime import ( File "/home/rohit/miniconda3/envs/vllm/lib/python3.10/site-packages/triton-2.1.0-py3.10-linux-x86_64.egg/triton/runtime/__init__.py", line 1, in <module> from .autotuner import (Autotuner, Config, Heuristics, OutOfResources, autotune, heuristics) File "/home/rohit/miniconda3/envs/vllm/lib/python3.10/site-packages/triton-2.1.0-py3.10-linux-x86_64.egg/triton/runtime/autotuner.py", line 7, in <module> from ..testing import do_bench, do_bench_cudagraph File "/home/rohit/miniconda3/envs/vllm/lib/python3.10/site-packages/triton-2.1.0-py3.10-linux-x86_64.egg/triton/testing.py", line 8, in <module> from . import language as tl File "/home/rohit/miniconda3/envs/vllm/lib/python3.10/site-packages/triton-2.1.0-py3.10-linux-x86_64.egg/triton/language/__init__.py", line 4, in <module> from . import math File "/home/rohit/miniconda3/envs/vllm/lib/python3.10/site-packages/triton-2.1.0-py3.10-linux-x86_64.egg/triton/language/math.py", line 5, in <module> from . import core File "/home/rohit/miniconda3/envs/vllm/lib/python3.10/site-packages/triton-2.1.0-py3.10-linux-x86_64.egg/triton/language/core.py", line 8, in <module> from .._C.libtriton.triton import ir ImportError: /home/rohit/miniconda3/envs/vllm/bin/../lib/libstdc++.so.6: version
GLIBCXX_3.4.30' not found (required by /home/rohit/miniconda3/envs/vllm/lib/python3.10/site-packages/triton-2.1.0-py3.10-linux-x86_64.egg/triton/_C/libtriton.so) ` help requested to resolve this.Aim: To get vllm working on a gentoo system.