Add BF16 README

bf16/README.md

@@ -0,0 +1,50 @@
+# shenwen-coderV2 BF16 GGUF
+
+This folder contains the **BF16 (BFloat16)** quantized version of [shenwenAI/shenwen-coderV2-Instruct](https://huggingface.co/shenwenAI/shenwen-coderV2-Instruct).
+
+## Model Info
+
+- **Architecture**: Qwen2 (0.5B parameters)
+- **Context Length**: 32,768 tokens
+- **Quantization**: BF16 (BFloat16)
+- **Original Precision**: bfloat16
+
+## Usage with swllm.cpp
+
+We recommend using our optimized [swllm.cpp](https://github.com/shenwenAI/swllm.cpp) for best performance:
+
+```bash
+# Build swllm.cpp (first time only)
+cd swllm.cpp
+mkdir build && cd build
+cmake .. && make -j4
+
+# Run inference
+./swllm -m shenwen-coderV2-BF16.gguf -p "Write a hello world in Python:" --top-p 0.9 -t 8
+```
+
+### Key Parameters
+
+| Parameter | Description | Recommended |
+|----------|-------------|-------------|
+| `-m` | Model path | Required |
+| `-t` | Thread count | CPU cores |
+| `-p` | Prompt | Your code prompt |
+| `--top-p` | Top-p sampling | 0.9 |
+| `-n` | Max tokens | 512-1024 |
+
+## Model Files
+
+| File | Size | Description |
+|------|------|-------------|
+| shenwen-coderV2-BF16.gguf | ~949 MB | Full BF16 precision |
+
+## Connect With Us
+
+[![GitHub](https://img.shields.io/badge/GitHub-shenwenAI-blue?style=flat&logo=github)](https://github.com/shenwenAI)
+[![Hugging Face](https://img.shields.io/badge/HuggingFace-shenwenAI-yellow?style=flat&logo=huggingface)](https://huggingface.co/shenwenAI)
+[![Twitter](https://img.shields.io/badge/Twitter-shenwenai-1DA1F2?style=flat&logo=twitter)](https://x.com/shenwenai)
+
+- GitHub: https://github.com/shenwenAI
+- Hugging Face: https://huggingface.co/shenwenAI
+- Twitter/X: https://x.com/shenwenai