JinnP/Qwen3-8B-Kernelbook-SFT-HF

TEXT GENERATIONPricing:Input $0.468 / Output $1.82Concurrent Unit Cost:1Model Size:8BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Aug 25, 2025License:apache-2.0Architecture:Transformer0.0K Open Weights Featherless Exclusive Cold

JinnP/Qwen3-8B-Kernelbook-SFT-HF is an 8.2 billion parameter causal language model, fine-tuned from Qwen3-8B by JinnP using Supervised Fine-Tuning (SFT) on the filtered KernelBook dataset. This model is specifically optimized for understanding, generating, and explaining content related to kernel internals, system-level programming, and operating system concepts. It excels in tasks such as kernel code generation, system debugging, and performance optimization recommendations, leveraging its specialized training data and 32768 token context length.

Loading preview...

Overview

JinnP/Qwen3-8B-Kernelbook-SFT-HF is an 8.2 billion parameter language model, fine-tuned from the Qwen3-8B base model. Developed by JinnP, this model underwent Supervised Fine-Tuning (SFT) using the SLIME framework, leveraging a highly curated subset of the KernelBook dataset, comprising 10,000 high-quality samples. The fine-tuning process specifically targeted kernel and system-level tasks, making it a specialized tool for developers and researchers working with operating systems.

Key Capabilities

  • Kernel Code Generation and Explanation: Proficient in generating and explaining code related to kernel internals.
  • System-Level Debugging: Aids in debugging scenarios within operating systems.
  • Performance Optimization: Provides recommendations for system performance improvements.
  • Operating System Concepts: Explains complex concepts such as memory management, process scheduling, device drivers, and file systems.

Training Details

The model's training focused on critical areas of system programming, including kernel internals, system calls, memory management, process scheduling, device drivers, I/O systems, file systems, networking stack, and performance optimization. This specialized training ensures its strong performance in these technical domains. The model is provided in a HuggingFace Transformers compatible format, ready for inference with various tools like vLLM and SGLang.

License

This model inherits the Apache 2.0 license from its base Qwen3-8B model.

Finetunes
1 models