[PR #25] [MERGED] Windows #1320

New issue

Closed

opened 2026-10-29 14:51:24 +00:00 by bsakaguchi · 0 comments

bsakaguchi commented

2026-10-29 14:51:24 +00:00

Owner

📋 Pull Request Information

Original PR: https://github.com/kvcache-ai/ktransformers/pull/25
Author: @Atream
Created: 8/8/2024
Status: ✅ Merged
Merged: 8/8/2024
Merged by: @UnicornChan

Base: main ← Head: windows

📝 Commits (3)

0a2fd52 support windows support q4_0 and q5_0 dequant on cpu Add CopyRight from pygguf(It was added before, but disappear after merge). Add some TODO in the code.
1d9d397 fix some bug in compile in linux
0e613b6 [fix] recover fp16 support

📊 Changes

33 files changed (+272 additions, -110 deletions)

View changed files

📝 ktransformers/ktransformers_ext/CMakeLists.txt (+19 -4)
📝 ktransformers/ktransformers_ext/cpu_backend/task_queue.h (+41 -3)
📝 ktransformers/ktransformers_ext/cuda/gptq_marlin/gptq_marlin.cu (+54 -19)
📝 ktransformers/ktransformers_ext/cuda/setup.py (+1 -1)
📝 ktransformers/ktransformers_ext/operators/llamafile/mlp.cpp (+5 -5)
📝 ktransformers/ktransformers_ext/operators/llamafile/moe.cpp (+18 -18)
📝 ktransformers/ktransformers_ext/operators/llamafile/moe.h (+1 -1)
📝 ktransformers/local_chat.py (+1 -1)
📝 ktransformers/models/custom_cache.py (+1 -0)
📝 ktransformers/operators/experts.py (+10 -3)
📝 ktransformers/optimize/optimize_rules/DeepSeek-V2-Chat.yaml (+1 -1)
📝 ktransformers/util/custom_gguf.py (+62 -17)
📝 ktransformers/util/utils.py (+8 -7)
📝 pyproject.toml (+2 -1)
📝 setup.py (+15 -2)
📝 third_party/llamafile/iqk_mul_mat.inc (+4 -3)
📝 third_party/llamafile/iqk_mul_mat_amd_avx2.cpp (+1 -1)
📝 third_party/llamafile/iqk_mul_mat_amd_zen4.cpp (+1 -1)
📝 third_party/llamafile/sgemm.cpp (+12 -7)
📝 third_party/llamafile/tinyblas_cpu.h (+1 -1)

...and 13 more files

📄 Description

support windows
support q4_0 and q5_0 dequant on cpu
Add CopyRight from pygguf(It was added before, but disappear after merge).
Add some TODO in the code.

_{🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.}

## 📋 Pull Request Information **Original PR:** https://github.com/kvcache-ai/ktransformers/pull/25 **Author:** [@Atream](https://github.com/Atream) **Created:** 8/8/2024 **Status:** ✅ Merged **Merged:** 8/8/2024 **Merged by:** [@UnicornChan](https://github.com/UnicornChan) **Base:** `main` ← **Head:** `windows` --- ### 📝 Commits (3) - [`0a2fd52`](https://github.com/kvcache-ai/ktransformers/commit/0a2fd52cea1f73cbbc222b998b0cb35c953bd8d2) support windows support q4_0 and q5_0 dequant on cpu Add CopyRight from pygguf(It was added before, but disappear after merge). Add some TODO in the code. - [`1d9d397`](https://github.com/kvcache-ai/ktransformers/commit/1d9d397525e07104f843030a567cb5271dee9df1) fix some bug in compile in linux - [`0e613b6`](https://github.com/kvcache-ai/ktransformers/commit/0e613b602d9e71d0754efcbd4875ca7c20a34c2b) [fix] recover fp16 support ### 📊 Changes **33 files changed** (+272 additions, -110 deletions) <details> <summary>View changed files</summary> 📝 `ktransformers/ktransformers_ext/CMakeLists.txt` (+19 -4) 📝 `ktransformers/ktransformers_ext/cpu_backend/task_queue.h` (+41 -3) 📝 `ktransformers/ktransformers_ext/cuda/gptq_marlin/gptq_marlin.cu` (+54 -19) 📝 `ktransformers/ktransformers_ext/cuda/setup.py` (+1 -1) 📝 `ktransformers/ktransformers_ext/operators/llamafile/mlp.cpp` (+5 -5) 📝 `ktransformers/ktransformers_ext/operators/llamafile/moe.cpp` (+18 -18) 📝 `ktransformers/ktransformers_ext/operators/llamafile/moe.h` (+1 -1) 📝 `ktransformers/local_chat.py` (+1 -1) 📝 `ktransformers/models/custom_cache.py` (+1 -0) 📝 `ktransformers/operators/experts.py` (+10 -3) 📝 `ktransformers/optimize/optimize_rules/DeepSeek-V2-Chat.yaml` (+1 -1) 📝 `ktransformers/util/custom_gguf.py` (+62 -17) 📝 `ktransformers/util/utils.py` (+8 -7) 📝 `pyproject.toml` (+2 -1) 📝 `setup.py` (+15 -2) 📝 `third_party/llamafile/iqk_mul_mat.inc` (+4 -3) 📝 `third_party/llamafile/iqk_mul_mat_amd_avx2.cpp` (+1 -1) 📝 `third_party/llamafile/iqk_mul_mat_amd_zen4.cpp` (+1 -1) 📝 `third_party/llamafile/sgemm.cpp` (+12 -7) 📝 `third_party/llamafile/tinyblas_cpu.h` (+1 -1) _...and 13 more files_ </details> ### 📄 Description support windows support q4_0 and q5_0 dequant on cpu Add CopyRight from pygguf(It was added before, but disappear after merge). Add some TODO in the code. --- <sub>🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.</sub>

bsakaguchi

2026-10-29 14:51:24 +00:00