Search results

small model and colab t4

enhancementNew feature or requestNew feature or request

Status: Open.

#2024 In kvcache-ai/ktransformers;

· kmnnmk212-source opened on May 26, 2026

完全使用hybird prefill时只加载冷门专家至内存，节约内存使用量

enhancementNew feature or requestNew feature or request

Status: Open.

#2023 In kvcache-ai/ktransformers;

· willbed opened on May 25, 2026

H200 部署 GLM 5.1

Status: Open.

#2019 In kvcache-ai/ktransformers;

· OldSixOne opened on May 21, 2026

swiglu_limit=10.0 is only supported on method='MXFP4'

Status: Open.

#2013 In kvcache-ai/ktransformers;

· mrgaolei opened on May 18, 2026

昇腾NPU在最新版本里面的支持

enhancementNew feature or requestNew feature or request

Status: Open.

#2012 In kvcache-ai/ktransformers;

· superxc opened on May 18, 2026

按照适配NPU的文档无法完成kt的安装

Status: Open.

#2011 In kvcache-ai/ktransformers;

· SongCQi opened on May 18, 2026

Crash when enable MTP on KT 0.6.2 DeepSeekV4Flash

Status: Open.

#2009 In kvcache-ai/ktransformers;

· mrgaolei opened on May 16, 2026

Ambiguous startup warning wrt. sglang / kt-kernel / moe integration

Status: Open.

#2008 In kvcache-ai/ktransformers;

· usrlocalben opened on May 16, 2026

Speculative/EAGLE - Supported?

Status: Open.

#2007 In kvcache-ai/ktransformers;

· usrlocalben opened on May 16, 2026

Discussion on the cost-effectiveness of AMX and FP8

Status: Open.

#2004 In kvcache-ai/ktransformers;

· mrgaolei opened on May 14, 2026

支持 SGLang 的 expert-parallel-size（EP）

enhancementNew feature or requestNew feature or request

Status: Open.

#2002 In kvcache-ai/ktransformers;

· willbed opened on May 13, 2026

[Help] Possible to run deepseek4-flash on ampere (3090) + threadripper 3970x (no AMX/AVX512) ?

Status: Open.

#1999 In kvcache-ai/ktransformers;

· sgmihai opened on May 11, 2026