| ---
|
| language:
|
| - en
|
| license: other
|
| license_name: deepseek
|
| library_name: transformers
|
| tags:
|
| - lattice
|
| - privacy
|
| - wrapped
|
| - large
|
| - code
|
| - instruct
|
| - reasoning
|
| base_model: deepseek-ai/DeepSeek-Coder-V2-Lite-Instruct
|
| ---
|
|
|
| # Lattice DeepSeek Coder V2 Lite
|
|
|
| **Privacy Tier:** `wrapped` | **Parameters:** 15.7B | **Context:** 163,840 tokens | **VRAM:** ~32GB
|
|
|
| DeepSeek Coder V2 Lite -- MoE architecture (2.4B active params from 15.7B total). Best open code model for its compute class. 338 programming languages. Fine-tune on proprietary codebases with DP guarantees.
|
|
|
| ## Privacy Guarantees
|
|
|
| | Feature | Status |
|
| |---|---|
|
| | Sandboxed training (no network egress) | Yes |
|
| | PII output guardrails | Yes |
|
| | Encrypted training logs | Yes |
|
| | Zero telemetry | Yes |
|
| | DP-SGD training support | Yes |
|
| | Privacy certificate on export | Yes |
|
|
|
| ## Quick Start
|
|
|
| ```bash
|
| pip install ltce
|
| ltce pull lattice-ai/deepseek-coder-v2-lite
|
| ltce train ./your-code --model deepseek-coder-v2-lite --epsilon 4.8 --method qlora
|
| ltce verify ./output/adapter
|
| ```
|
|
|
| ## Base Model
|
|
|
| [deepseek-ai/DeepSeek-Coder-V2-Lite-Instruct](https://huggingface.co/deepseek-ai/DeepSeek-Coder-V2-Lite-Instruct)
|
|
|
| ---
|
|
|
| Built with [Lattice](https://ltce.tech) -- Train private. Prove it. Share safely.
|
|
|
