https://github.com/crwsolutions/hfget
Download hugging face model
https://github.com/crwsolutions/hfget
Last synced: 3 days ago
JSON representation
Download hugging face model
- Host: GitHub
- URL: https://github.com/crwsolutions/hfget
- Owner: crwsolutions
- License: mit
- Created: 2026-04-15T19:33:13.000Z (about 2 months ago)
- Default Branch: main
- Last Pushed: 2026-04-15T21:23:29.000Z (about 2 months ago)
- Last Synced: 2026-04-15T23:25:12.979Z (about 2 months ago)
- Language: C#
- Size: 26.4 KB
- Stars: 0
- Watchers: 0
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# hfget
Download Hugging Face models (GGUF format) with parallel, resumable downloads.
## Features
- **Parallel downloads** - Download multiple files simultaneously with configurable thread count
- **Resumable downloads** - Automatically resume interrupted downloads
- **GGUF support** - Specifically designed for llama.cpp GGUF models
- **Smart file detection** - Automatically finds matching model and mmproj files
- **Progress tracking** - Real-time progress bars for all downloads
- **Llama.cpp integration** - Outputs ready-to-run llama-server command
## Installation
Download the latest release from the [GitHub releases page](https://github.com/crwsolutions/hfget/releases) or build from source:
```bash
dotnet build
```
## Usage
### Download a model
#### With Default command
```bash
hfget [targetDir]
```
#### Or with explicit command
```bash
hfget download [targetDir]
```
**Aliases:** `get`
**Arguments:**
- `` - Model identifier in format `author/model:variant`
- `[targetDir]` - Destination directory (default: current directory)
**Options:**
- `--token ` - Hugging Face API token (for private models)
- `--threads ` - Number of parallel download threads (default: 4)
- `-f, --folder` - Create a folder with the model name
### List model files
```bash
hfget list
```
**Arguments:**
- `` - Model identifier in format `author/model`
**Options:**
- `--token ` - Hugging Face API token (for private models)
## Examples
### Download a model to current directory (default)
```bash
hfget download bartowski/Llama-3.2-1B-Instruct-GGUF:q4_k_m
```
This will:
1. Search for a `.gguf` file containing `q4_k_m` in the filename
2. Look for an associated `mmproj` file (for vision models)
3. Download matching files to the current directory
4. Print a command to run with `llama-server.exe`
### Download a model to a specific directory
```bash
hfget download bartowski/Llama-3.2-1B-Instruct-GGUF:q4_k_m ./models
```
### Download with custom thread count
```bash
hfget get "mistralai/Mistral-7B-Instruct-v0.3:q4_k_m" --threads 8 ./models
```
### Download a private model
```bash
hfget download --token hf_XXXXXXXXXX private-user/private-model:q5_k_m
```
### Download with folder organization
```bash
hfget download -f unsloth/Qwen3.5-2B-GGUF:IQ2_XXS ./models
```
This will:
1. Search for a `.gguf` file containing `IQ2_XXS` in the filename
2. Look for an associated `mmproj` file (for vision models)
3. Create a folder `models/unsloth_Qwen3.5-2B-GGUF_IQ2_XXS`
4. Download files with their original names inside the folder
5. Print a command to run with `llama-server.exe`
### List available files in a model
```bash
hfget list bartowski/Llama-3.2-1B-Instruct-GGUF
```
Output shows all files with their types (GGUF, mmproj, other) and sizes.
After downloading, hfget displays a ready-to-run command:
```
Run command:
.\llama-server.exe -m "./models/model.gguf" -mm "./models/mmproj.gguf"
```
## How It Works
1. **Parse model identifier** - Extracts repository name and variant from input
2. **Fetch file list** - Queries Hugging Face API for repository files
3. **Find matching files** - Searches for `.gguf` files containing the variant and `mmproj` files
4. **Download files** - Downloads files in parallel with progress tracking
5. **Generate command** - Outputs a ready-to-run llama.cpp server command
## License
MIT License - see [LICENSE](LICENSE) file for details.