GithubStar/ktransformers
1
0
Fork 0
mirror of https://github.com/kvcache-ai/ktransformers.git synced 2026-10-16 21:01:14 +08:00
Code Issues Packages Projects Releases Wiki Activity

fix(kt-kernel): cap compressed-tensors below the torch>=2.10 releases (#2116)

Browse Source 
Signed-off-by: Anai-Guo <antai12232931@outlook.com>
This commit is contained in:
Tai An
2026-09-22 22:07:33 -07:00
committed by GitHub
parent f7607c0f66
commit c40722bf04
1 changed files with 5 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files
kt-kernel/pyproject.toml
+5 -1
View File
@@ -44,7 +44,11 @@ sglang = [
"sglang-kt",
]
convert = [
"compressed-tensors>=0.7.0",
# compressed-tensors 0.16.0 moved its offload stack onto torch.accelerator
# and requires torch>=2.10.0, but this package pins torch==2.9.1. Installing
# the newer release against that torch makes convert_gpu_weights.py fail with
# "module 'torch.accelerator' has no attribute 'get_memory_info'".
"compressed-tensors>=0.7.0,<0.16.0",
]
test = [
"pytest>=7.0.0",