[GH-ISSUE #23] Failed to build wheels by myself #13

New issue

Closed

opened 2026-10-29 14:32:39 +00:00 by bsakaguchi · 2 comments

bsakaguchi commented

2026-10-29 14:32:39 +00:00

Owner

Originally created by @RoacherM on GitHub (Aug 6, 2024).
Original GitHub issue: https://github.com/kvcache-ai/ktransformers/issues/23

I tried to compile this library myself, but ultimately failed. Do you have a pre-compiled .whl file available? My current machine environment is CUDA 12.2, PyTorch 2.3+cu121, and the GPU model is H100.

` Running command Building wheel for ktransformers (pyproject.toml)
Using native cpu instruct
running bdist_wheel
Using native cpu instruct
Guessing wheel URL: https://github.com/kvcache-ai/ktransformers/releases/download/v0.1.1/ktransformers-0.1.1+cu122torch23fancy-cp310-cp310-linux_x86_64.whl
Raw wheel path /tmp/pip-wheel-bxz7xsx4/.tmp-1zlb72in/ktransformers-0.1.1-cp310-cp310-linux_x86_64.whl
error: [Errno 18] Invalid cross-device link: 'ktransformers-0.1.1+cu122torch23fancy-cp310-cp310-linux_x86_64.whl' -> '/tmp/pip-wheel-bxz7xsx4/.tmp-1zlb72in/ktransformers-0.1.1-cp310-cp310-linux_x86_64.whl'
error: subprocess-exited-with-error

× Building wheel for ktransformers (pyproject.toml) did not run successfully.
│ exit code: 1
╰─> See above for output.

note: This error originates from a subprocess, and is likely not a problem with pip.
full command: /mnt/disk1/envs/default/miniconda3/envs/sglang/bin/python /mnt/disk1/envs/default/miniconda3/envs/sglang/lib/python3.10/site-packages/pip/_vendor/pyproject_hooks/_in_process/_in_process.py build_wheel /tmp/tmpb1_b9a00
cwd: /mnt/disk1/wy/ktransformers
Building wheel for ktransformers (pyproject.toml) ... error
ERROR: Failed building wheel for ktransformers
Failed to build ktransformers
ERROR: ERROR: Failed to build installable wheels for some pyproject.toml based projects (ktransformers)`

Originally created by @RoacherM on GitHub (Aug 6, 2024). Original GitHub issue: https://github.com/kvcache-ai/ktransformers/issues/23 I tried to compile this library myself, but ultimately failed. Do you have a pre-compiled .whl file available? My current machine environment is CUDA 12.2, PyTorch 2.3+cu121, and the GPU model is H100. ` Running command Building wheel for ktransformers (pyproject.toml) Using native cpu instruct running bdist_wheel Using native cpu instruct Guessing wheel URL: https://github.com/kvcache-ai/ktransformers/releases/download/v0.1.1/ktransformers-0.1.1+cu122torch23fancy-cp310-cp310-linux_x86_64.whl Raw wheel path /tmp/pip-wheel-bxz7xsx4/.tmp-1zlb72in/ktransformers-0.1.1-cp310-cp310-linux_x86_64.whl error: [Errno 18] Invalid cross-device link: 'ktransformers-0.1.1+cu122torch23fancy-cp310-cp310-linux_x86_64.whl' -> '/tmp/pip-wheel-bxz7xsx4/.tmp-1zlb72in/ktransformers-0.1.1-cp310-cp310-linux_x86_64.whl' error: subprocess-exited-with-error × Building wheel for ktransformers (pyproject.toml) did not run successfully. │ exit code: 1 ╰─> See above for output. note: This error originates from a subprocess, and is likely not a problem with pip. full command: /mnt/disk1/envs/default/miniconda3/envs/sglang/bin/python /mnt/disk1/envs/default/miniconda3/envs/sglang/lib/python3.10/site-packages/pip/_vendor/pyproject_hooks/_in_process/_in_process.py build_wheel /tmp/tmpb1_b9a00 cwd: /mnt/disk1/wy/ktransformers Building wheel for ktransformers (pyproject.toml) ... error ERROR: Failed building wheel for ktransformers Failed to build ktransformers ERROR: ERROR: Failed to build installable wheels for some pyproject.toml based projects (ktransformers)`

bsakaguchi closed this issue

2026-10-29 14:32:40 +00:00

bsakaguchi commented

2026-10-29 14:32:40 +00:00

Author

Owner

@UnicornChan commented on GitHub (Aug 6, 2024):

The pre-compiled package can be downloaded from the URL mentioned in the error log https://github.com/kvcache-ai/ktransformers/releases/download/v0.1.1/ktransformers-0.1.1+cu122torch23fancy-cp310-cp310-linux_x86_64.whl
However, the pre-compiled package was built with TORCH_CUDA_ARCH_LIST="8.0;8.6;8.7;8.9" and does not include "9.0" which is the arch for H100. I am not sure if the pre-compiled package will run on H100.
I suggest compile it, For compile this library, use

KTRANSFORMERS_FORCE_BUILD=TRUE pip install . --no-build-isolation

instead of using the following command pip install . --no-build-isolation in install.sh
I will correct it right away

@UnicornChan commented on GitHub (Aug 6, 2024): The pre-compiled package can be downloaded from the URL mentioned in the error log [https://github.com/kvcache-ai/ktransformers/releases/download/v0.1.1/ktransformers-0.1.1+cu122torch23fancy-cp310-cp310-linux_x86_64.whl](https://github.com/kvcache-ai/ktransformers/releases/download/v0.1.1/ktransformers-0.1.1+cu122torch23fancy-cp310-cp310-linux_x86_64.whl) However, the pre-compiled package was built with TORCH_CUDA_ARCH_LIST="8.0;8.6;8.7;8.9" and does not include "9.0" which is the arch for H100. I am not sure if the pre-compiled package will run on H100. I suggest compile it, For compile this library, use ``` KTRANSFORMERS_FORCE_BUILD=TRUE pip install . --no-build-isolation ``` instead of using the following command ```pip install . --no-build-isolation``` in [install.sh](https://github.com/kvcache-ai/ktransformers/blob/v0.1.1/install.sh) I will correct it right away

bsakaguchi commented

2026-10-29 14:32:40 +00:00

Author

Owner

@chuangzhidan commented on GitHub (Feb 20, 2025):

The pre-compiled package can be downloaded from the URL mentioned in the error log https://github.com/kvcache-ai/ktransformers/releases/download/v0.1.1/ktransformers-0.1.1+cu122torch23fancy-cp310-cp310-linux_x86_64.whl However, the pre-compiled package was built with TORCH_CUDA_ARCH_LIST="8.0;8.6;8.7;8.9" and does not include "9.0" which is the arch for H100. I am not sure if the pre-compiled package will run on H100. I suggest compile it, For compile this library, use
KTRANSFORMERS_FORCE_BUILD=TRUE pip install . --no-build-isolation
instead of using the following command pip install . --no-build-isolation in install.sh I will correct it right away

KTRANSFORMERS_FORCE_BUILD=TRUE pip install . --no-build-isolation doesn't work for me , mine is A800. this installing thing is really not friendly (a bitch), i have been working on this for days!

@chuangzhidan commented on GitHub (Feb 20, 2025): > The pre-compiled package can be downloaded from the URL mentioned in the error log https://github.com/kvcache-ai/ktransformers/releases/download/v0.1.1/ktransformers-0.1.1+cu122torch23fancy-cp310-cp310-linux_x86_64.whl However, the pre-compiled package was built with TORCH_CUDA_ARCH_LIST="8.0;8.6;8.7;8.9" and does not include "9.0" which is the arch for H100. I am not sure if the pre-compiled package will run on H100. I suggest compile it, For compile this library, use > > ``` > KTRANSFORMERS_FORCE_BUILD=TRUE pip install . --no-build-isolation > ``` > > instead of using the following command `pip install . --no-build-isolation` in [install.sh](https://github.com/kvcache-ai/ktransformers/blob/v0.1.1/install.sh) I will correct it right away **KTRANSFORMERS_FORCE_BUILD=TRUE pip install . --no-build-isolation doesn't work for me , mine is A800. this installing thing is really not friendly (a bitch), i have been working on this for days!**

bsakaguchi referenced this issue

2026-10-29 14:37:56 +00:00

[GH-ISSUE #465] /pytorch/aten/src/ATen/native/cuda/TensorCompare.cu:110: _assert_async_cuda_kernel: block: [0,0,0], thread: [0,0,0] Assertion probability tensor contains either inf, nan or element < 0 failed. #327

bsakaguchi referenced this issue

2026-10-29 14:42:22 +00:00

[GH-ISSUE #903] Rocm run deepseek r1 671b q4 failed #646

bsakaguchi referenced this issue

2026-10-29 14:45:29 +00:00

[GH-ISSUE #1269] [Question] benchmark with evalscope failed #859

bsakaguchi referenced this issue

2026-10-29 14:45:38 +00:00

[GH-ISSUE #1286] [Bug] inference_context.v_cache[0][i] IndexError: list index out of range #871