kvcache-ai / kvcache-ai/ktransformers
TypeError: KTMoEWrapper.__new__() got an unexpected keyword argument 'num_gpu_experts'
- Dominant language
- Python
- Stars
- 19.5k
- Forks
- 1.6k
- Avg merge
- 19h 32m
- Merged PRs (30d)
- 27
Description
### Reminder
- [x] I have read the above rules and searched the existing issues.
### System Info
kt_kernel version: 0.5.1 from pypi
sglang version: 0.5.8 from pypi
OS: RHEL 9.6
CPU: Intel Xeon GOLD 6530
GPU: L20*4
CUDA version: 12.9
python version: 3.12.7
### Reproduction
I run sglang.launch_server command and the error will be encountered, I think the reason is [the `CPU-GPU experts sched`](https://github.com/kvcache-ai/ktransformers/commit/027832c590c1b08e228ed60987eb22ae89283602) feature remove the `num_gpu_experts` argument in KTMoEWrapper, I found that your forked branch of sglang fixed this. Are there any plans to upload this fix to the main branch of sglang?
### Others
_No response_
Contributor guide
Research direction
Start with the reported sglang.launch_server failure and the KTMoEWrapper constructor error, then compare the referenced CPU-GPU experts sched commit with the forked sglang change mentioned in the report. Confirm which versions still pass num_gpu_experts and define done as restoring compatibility or documenting the supported version relationship.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python
- Domain
- ai
- Issue type
- Bug
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 35/100