| --- |
| title: Mistral Nemo Alpaca Finetune Demo |
| emoji: 🚀 |
| colorFrom: purple |
| colorTo: purple |
| sdk: gradio |
| sdk_version: 6.26.0 |
| app_file: app.py |
| pinned: false |
| hardware: t4-medium |
| license: apache-2.0 |
| --- |
| |
| # Mistral Nemo Alpaca Finetune Demo |
|
|
| This Space provides an interactive chat demo for the [TensorVizion/mistral-nemo-alpaca-finetune](https://huggingface.co/TensorVizion/mistral-nemo-alpaca-finetune) model. |
|
|
| ## Model Details |
| - **Base Model:** Mistral Nemo (12B) |
| - **Finetuned by:** TensorVizion |
| - **Training:** Fine-tuned 2x faster with Unsloth and Hugging Face's TRL library on Alpaca-style instructions. |
| - **Precision:** Loaded in 4-bit quantization (NF4) for efficient inference on consumer-grade GPUs. |
|
|
| ## Hardware Requirements |
| This Space is configured to run on a `t4-medium` GPU (16GB VRAM). The 4-bit quantization ensures the 12B parameter model fits comfortably within this memory limit (~8GB VRAM usage). If you experience out-of-memory errors or want faster generation, consider upgrading to an `a10g-small` hardware tier in the Space settings. |
