deepseekmirror/DeepSeek-V3
1
0
Fork 0
mirror of https://github.com/deepseek-ai/DeepSeek-V3.git synced 2025-07-04 23:41:37 -04:00
Code Issues Actions 1 Packages Projects Releases Wiki Activity

docs: Improved README, add additional references

Browse Source
This commit is contained in:
Triex 2025-06-05 04:08:59 +10:00
parent 9aedaae1d5
commit 9c25e23773
1 changed files with 22 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

23
README.md
View File

@ -30,7 +30,7 @@ Current LLM inference is dominated by Python/PyTorch, which introduces:
- **Complex deployment** with heavy runtimes - **Complex deployment** with heavy runtimes
- **Platform lock-in** due to dependency complexity - **Platform lock-in** due to dependency complexity
## The Zig Advantage ## Why Zig?
**Performance**: Zero-cost abstractions, compile-time optimization, direct hardware access **Performance**: Zero-cost abstractions, compile-time optimization, direct hardware access
**Simplicity**: Single static binary, no runtime dependencies, cross-compilation built-in **Simplicity**: Single static binary, no runtime dependencies, cross-compilation built-in
@ -180,6 +180,8 @@ This is an ambitious project that would benefit from expertise in:
- **ML model implementation** - **ML model implementation**
- **Web server development** - **Web server development**
- **Performance optimization** - **Performance optimization**
- **Hardware-software co-design**
- **Novel inference techniques** (Speculative decoding, quantization)
## Project Timeline ## Project Timeline
@ -188,12 +190,31 @@ This is an ambitious project that would benefit from expertise in:
- Backend optimization and web API - Backend optimization and web API
- Testing, benchmarking, deployment tools - Testing, benchmarking, deployment tools
## Key Questions
**Q: Why not just optimize PyTorch?**
A: PyTorch's Python overhead and GC pauses are fundamental limitations. Zig offers zero-cost abstractions, superior error handling, and deterministic performance.
**Q: How will this compare to llama.cpp?**
A: Similar performance goals, but with built-in web API, better memory management, and focus on DeepSeek V3's specific MoE architecture.
**Q: What about ONNX/TensorRT/ZML etc?**
A: Those are inference runtimes, not development frameworks / LLM frameworks. This project enables rapid iteration and custom optimization for research.
---
**Status**: 🎯 Seeking feedback & idea expansion
**Vision**: Foundation for advanced AI reasoning research
## References ## References
- [DeepSeek V3 Paper](https://arxiv.org/abs/2412.19437) - Original model architecture - [DeepSeek V3 Paper](https://arxiv.org/abs/2412.19437) - Original model architecture
- [Zig Language](https://ziglang.org/) - Language documentation - [Zig Language](https://ziglang.org/) - Language documentation
- [Awesome Zig](https://github.com/C-BJ/awesome-zig) - Community resources - [Awesome Zig](https://github.com/C-BJ/awesome-zig) - Community resources
- [Zig Patterns](https://github.com/SuperAuguste/zig-patterns) - Common idioms - [Zig Patterns](https://github.com/SuperAuguste/zig-patterns) - Common idioms
- [ZML](https://github.com/zml/zml) - Zig Inference Stack
- [LLaMA.cpp](https://github.com/ggerganov/llama.cpp) - C++ Inference Engine
- [DeepZig Consciousness](https://github.com/Triex/DeepZig-Consciousness) - Research goal/end game
--- ---