File size: 2,950 Bytes
de3b009 3ab0dab 7d8b0d7 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 | 
NanoCoder is a Fill-in-the-Middle (FIM) language model specifically designed for React frontend development and coding assistance. It helps users with intelligent code autocompletion and context-aware generation. The model was fine-tuned using Unsloth on the Qwen 3 0.6B base model, leveraging a high-quality FIM dataset curated from GitHub repositories to enhance coding capabilities and developer productivity.
## 🧠 Datasets
We trained **NanoCoder** using a **high-quality Fill-in-the-Middle (FIM)** dataset curated from GitHub repositories:
[**srisree/nextjs_typescript_fim_dataset**](https://huggingface.co/datasets/srisree/nextjs_typescript_fim_dataset) on Hugging Face.
This dataset focuses on **React/Next.js** and **TypeScript** projects, providing rich, real-world coding examples that help the model understand frontend architecture, component composition, and React ecosystem patterns.
By leveraging this dataset, **NanoCoder** learns to:
- Predict and fill missing code intelligently using FIM objectives.
- Understand React component structures and TypeScript typing patterns.
- Generate clean, production-grade frontend code snippets.
## ⚙️ FIM Training Colab Script
We’re preparing an interactive **Google Colab notebook** for reproducing the **Fill-in-the-Middle (FIM)** fine-tuning process used to train **NanoCoder** with **Unsloth** on the **Qwen 3 0.6B** base model.
The Colab script will include:
- ✅ Environment setup with Unsloth and Qwen 3 0.6B
- ✅ Loading and preprocessing the [Next.js TypeScript FIM Dataset](https://huggingface.co/datasets/srisree/nextjs_typescript_fim_dataset)
- ✅ Training configuration (LoRA, batch size, sequence length, etc.)
- ✅ Evaluation and inference examples
🚀 **Coming soon...** Stay tuned for the full release!
---
# ⚙️ Setup and Run NanoCoder Locally with Ollama in VS Code
> Step-by-step guide to install, configure, and use **NanoCoder** for intelligent React frontend code completion with the **Continue** VS Code extension.
---
## 🧠 Prerequisites
Before getting started, ensure you have the following installed:
- [VS Code](https://code.visualstudio.com/)
- [Ollama](https://ollama.ai) (latest version)
- [Continue extension](https://marketplace.visualstudio.com/items?itemName=Continue.continue)
- A system with at least **8GB RAM** (recommended for 0.6B models)
---
## 🧩 Step 1: Install Ollama
If you haven’t already, download and install **Ollama**:
- macOS / Linux / Windows: [https://ollama.ai/download](https://ollama.ai/download)
Once installed, open your terminal and verify the installation.
## 💾 Step 2: Pull NanoCoder Model
>ollama pull srisree/nanocoder
## ⚡ Step 3: Run NanoCoder with Ollama
Once downloaded, you can test NanoCoder directly in the terminal:
>ollama run nanocoder
Read more [Continue Docs](https://docs.continue.dev/customize/deep-dives/autocomplete) |