Upload README.md with huggingface_hub
Browse files
README.md
ADDED
|
@@ -0,0 +1,46 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: cc-by-4.0
|
| 3 |
+
base_model: Cloudflare/clef-flash
|
| 4 |
+
tags:
|
| 5 |
+
- gguf
|
| 6 |
+
- rocm
|
| 7 |
+
- rocmfpx
|
| 8 |
+
- amd
|
| 9 |
+
- llama.cpp
|
| 10 |
+
---
|
| 11 |
+
|
| 12 |
+
# cloudflare-clef-flash-ROCmFPX - ROCmFPX Quantized
|
| 13 |
+
|
| 14 |
+
This repository contains AMD ROCmFPX quantized GGUF weights converted from [Cloudflare/clef-flash](https://huggingface.co/Cloudflare/clef-flash) using [charlie12345/ROCmFPX](https://github.com/charlie12345/ROCmFPX).
|
| 15 |
+
|
| 16 |
+
## Available Files
|
| 17 |
+
|
| 18 |
+
| File | Size | Format |
|
| 19 |
+
| --- | --- | --- |
|
| 20 |
+
| `clef-flash-Q4_0_ROCMFP4_FAST.gguf` | 4.44 GB | Q4_0_ROCMFP4_FAST |
|
| 21 |
+
| `clef-flash-Q4_0_ROCMFP4.gguf` | 5.47 GB | Q4_0_ROCMFP4 |
|
| 22 |
+
| `clef-flash-Q4_0_ROCMFP4_COHERENT.gguf` | 4.95 GB | Q4_0_ROCMFP4_COHERENT |
|
| 23 |
+
| `clef-flash-Q4_0_ROCMFP4_FAST_COHERENT.gguf` | 4.72 GB | Q4_0_ROCMFP4_FAST_COHERENT |
|
| 24 |
+
| `clef-flash-Q4_0_ROCMFP4_LEAN.gguf` | 4.82 GB | Q4_0_ROCMFP4_LEAN |
|
| 25 |
+
| `clef-flash-Q4_0_ROCMFP4_STRIX.gguf` | 4.74 GB | Q4_0_ROCMFP4_STRIX |
|
| 26 |
+
| `clef-flash-Q4_0_ROCMFP4_STRIX_LEAN.gguf` | 4.62 GB | Q4_0_ROCMFP4_STRIX_LEAN |
|
| 27 |
+
| `clef-flash-Q2_0_ROCMFPX.gguf` | 3.10 GB | Q2_0_ROCMFPX |
|
| 28 |
+
| `clef-flash-Q3_0_ROCMFPX.gguf` | 4.60 GB | Q3_0_ROCMFPX |
|
| 29 |
+
| `clef-flash-Q6_0_ROCMFPX.gguf` | 6.86 GB | Q6_0_ROCMFPX |
|
| 30 |
+
| `clef-flash-Q8_0_ROCMFPX.gguf` | 8.61 GB | Q8_0_ROCMFPX |
|
| 31 |
+
| `clef-flash-Q8_0_ROCMFPX_AGENT.gguf` | 8.77 GB | Q8_0_ROCMFPX_AGENT |
|
| 32 |
+
|
| 33 |
+
|
| 34 |
+
## Usage Instructions with ROCmFPX llama.cpp
|
| 35 |
+
|
| 36 |
+
```bash
|
| 37 |
+
git clone --depth 1 [https://github.com/charlie12345/ROCmFPX.git](https://github.com/charlie12345/ROCmFPX.git)
|
| 38 |
+
cd ROCmFPX
|
| 39 |
+
cmake -B build -G Ninja -DCMAKE_BUILD_TYPE=Release -DGGML_CUDA=OFF -DGGML_NATIVE=ON
|
| 40 |
+
cmake --build build --target llama-cli -j
|
| 41 |
+
|
| 42 |
+
./build/bin/llama-cli -m clef-flash-Q4_0_ROCMFP4_FAST.gguf -p "You are a helpful assistant. Hello!" -n 128
|
| 43 |
+
```
|
| 44 |
+
|
| 45 |
+
## License
|
| 46 |
+
This model is distributed under the [Creative Commons Attribution 4.0 International (CC BY 4.0)](https://creativecommons.org/licenses/by/4.0/) license.
|