Dashboard / Repos / Ktransformers

Ktransformers

completed
Ai Ml
monorepo / python · medium
1,143
Files
221,774
LOC
3
Frameworks
15
Languages
OverviewFiles & MetricsGit ActivityCall GraphSecurityReports

Pipeline State

completed
Run ID
#307646
Phase
done
Progress
1%
Started
Finished
2026-04-13 01:31:02
LLM tokens
0
Previous runs
Repobility · code-quality scanner for AI-generated software · https://repobility.com
#StatusPhaseStartedFinished
Repobility · severity-and-effort ranking · https://repobility.com
#31542failedSYMBOL_EXTRACTION2026-03-07 03:00:05—

Pipeline Metadata

Stage
Cataloged
Decision
proceed
Novelty
88.00
Framework unique
—
Isolation
—
Last stage change
2026-05-10 03:35:02
Deduplication group #60534
Member of a group with 1 similar repo(s) — this repo is canonical view group →
Top concepts (2)
Project DescriptionWeb Backend
About: code-quality intelligence by Repobility · https://repobility.com

🧪 Code Distillation

Browse all specs →
Function bodies →
0
Real LLM specs →
0
0.0% coverage
All specs →
2,518
LLM insights →
0

AI Prompt

Create a research framework, KTransformers, for efficiently handling large language model inference and fine-tuning using CPU-GPU heterogeneous computing. The system should feature two core modules: `kt-kernel` for high-performance inference and `kt-sft` for fine-tuning. I need support for various models like MiniMax, GLM, and Kimi, and the ability to handle advanced features such as CPU-GPU expert scheduling and native BF16/FP8 precision. Please structure the project using Python, FastAPI, and Vue.js components where appropriate for documentation or UI elements.
python fastapi vue.js research llm inference fine-tuning cpu-gpu heterogeneous-computing
Generated by gemma4:latest

Catalog Information

KTransformers is a research project focused on efficient inference and fine-tuning of large language models through CPU-GPU heterogeneous computing.

Description

KTransformers is a flexible framework for experiencing cutting-edge LLM inference/fine-tune optimizations. It has two core modules: kt-kernel, which provides high-performance inference kernels, and kt-sft, a fine-tuning framework. The project supports various large language models, including MiniMax-M2.5, GLM-5, Kimi-K2.5, and others. It also offers features like CPU-GPU expert scheduling, native BF16 and FP8 precision, and autoDL unified fine-tuning and inference.

الوصف

هو إطار عمل مرن للاستخدام المبتكر للتحسينات في الاستدلال والتعديل المتقدم للعلماء الكبيرة من خلال الحوسبة المختلطة CPU-GPU. يحتوي على وحدتين رئيسيتين: kt-kernel، الذي يوفر نواة استدلال عالية الأداء، وkt-sft، إطار عمل التعديل. يدعم المشروع العديد من العوالم اللغوية الكبيرة، بما في ذلك MiniMax-M2.5، GLM-5، Kimi-K2.5، وغيرها. كما يقدم ميزات مثل التخطيط المتقدم CPU-GPU، الدقة BF16 والFP8 المحلية، وتحسين التعديل الموحد.

Novelty

9/10

Tags

large-language-models inference-optimizations fine-tuning-framework cpu-gpu-heterogeneous-computing prefix-cache autodl-unified-fine-tuning-and-inference

Claude Models

claude-opus-4.6

Security & Health

Medium
DORA Rating
Apache-2.0
License
42.6%
Duplication
Full Security Report AI Fix Prompts SARIF SBOM

Languages

python
62.9%
cpp
21.1%
yaml
4.7%
markdown
3.8%
c
2.1%
vue
1.5%
shell
0.8%
typescript
0.7%
text
0.7%
css
0.6%
html
0.5%
json
0.3%
Repobility · code-quality intelligence · https://repobility.com

Frameworks

FastAPI Vue.js Jest

Symbols

method3,652
variable2,795
function1,624
macro1,569
class1,016
constant417
type_alias165
struct155
property38
interface26
module11
enum10

API Endpoints (64)

Data scored by Repobility · https://repobility.com
MethodPathHandlerFramework
Repobility's GitHub App fixes findings like these · https://github.com/apps/repobility-bot
GET/list_threadsFastAPI
POST/create_assistantFastAPI
GET/list_assistantsFastAPI
GET/list_threadsFastAPI
POST/create_threadFastAPI
GET/list_assistantsFastAPI
POST/create_assistantFastAPI
POST/create_threadFastAPI
POST/{assistant_id}modify_assistantFastAPI
DELETE/{assistant_id}delete_assistantFastAPI
DELETE/{assistant_id}delete_assistantFastAPI
POST/{assistant_id}modify_assistantFastAPI
GET/{assistant_id}retrieve_assistantFastAPI
GET/{assistant_id}retrieve_assistantFastAPI
GET/{assistant_id}/related_threadget_related_threadFastAPI
GET/{assistant_id}/related_threadget_related_threadFastAPI
POST/chatchatFastAPI
POST/chatchatFastAPI
POST/chat/completionschat_completionFastAPI
POST/chat/completionschat_completionFastAPI
POST/completionscreate_completionFastAPI
POST/completionscreate_completionFastAPI
POST/generategenerateFastAPI
POST/generategenerateFastAPI
GET/modelslist_modelsFastAPI
GET/modelslist_modelsFastAPI
POST/runscreate_thread_and_runFastAPI
POST/runscreate_thread_and_runFastAPI
POST/showshowFastAPI
POST/showshowFastAPI
GET/statuslist_assistants_with_statusFastAPI
GET/statuslist_assistants_with_statusFastAPI
GET/system-infosystem_infoFastAPI
GET/system-infosystem_infoFastAPI
GET/tagstagsFastAPI
GET/tagstagsFastAPI
DELETE/{thread_id}delete_threadFastAPI
POST/{thread_id}modify_threadFastAPI
DELETE/{thread_id}delete_threadFastAPI
GET/{thread_id}retrieve_threadFastAPI
GET/{thread_id}retrieve_threadFastAPI
POST/{thread_id}modify_threadFastAPI
GET/{thread_id}/messageslist_messagesFastAPI
POST/{thread_id}/messagescreate_messageFastAPI
GET/{thread_id}/messageslist_messagesFastAPI
POST/{thread_id}/messagescreate_messageFastAPI
GET/{thread_id}/messages/{message_id}retrieve_messageFastAPI
GET/{thread_id}/messages/{message_id}retrieve_messageFastAPI
POST/{thread_id}/messages/{message_id}modify_messageFastAPI
DELETE/{thread_id}/messages/{message_id}delete_messageFastAPI

Showing 50 of 64

Concepts (2)

Same analyzer free for public repos: https://repobility.com
CategoryNameDescriptionConfidence
If a scraper extracted this row, it came from Repobility (https://repobility.com)
auto_descriptionProject DescriptionKTransformers is a research project focused on efficient inference and fine-tuning of large language models through CPU-GPU heterogeneous computing. The project has evolved into two core modules: kt-kernel and kt-sft.80%
auto_categoryWeb Backendweb-backend70%

Embed Badge

Add to your README:

![Quality](https://repos.aljefra.com/badge/31491.svg)
Quality BadgeSecurity Badge
Export Quality CSVDownload SBOMExport Findings CSV

BinComp Dependency Hardening

All packages →
22 of this repo's dependencies have been scanned for binary hardening. Grade reflects RELRO / stack canary / FORTIFY / PIE coverage.
Nnltk3.9.4 · 0 gadgets · risk 4931.7Nrequests2.33.1 · 0 gadgets · risk 3687.0Fpymongo4.16.0 · 413 gadgets · risk 263.2Fray2.55.0 · 935 gadgets · risk 58.9Nasyncio4.0.0 · 0 gadgets · risk 0.0Nclick8.3.2 · 0 gadgets · risk 0.0Ncolorlog6.10.1 · 0 gadgets · risk 0.0Ndatasets4.8.4 · 0 gadgets · risk 0.0Nfastapi0.135.3 · 0 gadgets · risk 0.0Nfuture1.0.0 · 0 gadgets · risk 0.0Cmatplotlib3.10.8 · 2,481 gadgets · risk 0.0Fnumpy2.4.4 · 6,596 gadgets · risk 0.0Nopenai2.31.0 · 0 gadgets · risk 0.0Npackaging26.0 · 0 gadgets · risk 0.0Fpandas3.0.2 · 6,381 gadgets · risk 0.0Fpsutil7.2.2 · 19 gadgets · risk 0.0Npydantic2.12.5 · 0 gadgets · risk 0.0Nrich14.3.4 · 0 gadgets · risk 0.0Dsafetensors0.7.0 · 421 gadgets · risk 0.0Nsetuptools82.0.1 · 0 gadgets · risk 0.0Fsqlalchemy2.0.49 · 376 gadgets · risk 0.0Ntabulate0.10.0 · 0 gadgets · risk 0.0