|
|
--- |
|
|
license: apache-2.0 |
|
|
--- |
|
|
Here, we provide AWQ-quantized versions of the most popular NVILA models. These files help you seamlessly deploy TinyChat to unlock the full potential of NVILA and your hardware. |
|
|
|
|
|
One-command demo to chat with quantized NVILA models via [llm-awq](https://github.com/mit-han-lab/llm-awq/tree/main/tinychat#:~:text=TinyChat%20support%20NVILA) (NVILA-8B as an example): |
|
|
|
|
|
```bash |
|
|
cd llm-awq/tinychat |
|
|
python nvila_demo.py --model-path PATH/TO/NVILA \ |
|
|
--quant_path NVILA-8B-w4-g128-awq-v2.pt \ |
|
|
--media PATH/TO/ANY/IMAGES/VIDEOS \ |
|
|
--act_scale_path NVILA-8B-VT-smooth-scale.pt \ |
|
|
--all --chunk --model_type nvila |
|
|
``` |
|
|
|
|
|
This command will download the quantized NVILA model and run a chat demo. If you’ve already downloaded the files, simply set the path to your local copies. |
|
|
|
