OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Models

Camoscio

Camoscio is an Italian instruction-tuned language model based on LLaMA and Stanford Alpaca, trained using low-rank adaptation. Its public repository provides model weights, an Italian instruction dataset, code to reproduce training, a demo notebook and guidance for fine-tuning.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Camoscio — OpenViST project profile in AI Models.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use Camoscio when you need a research model and repository materials for experimenting with Italian instruction-following generation or task-specific fine-tuning.

Who it is for

  • Researchers exploring Italian instruction-tuned language models.
  • Developers prototyping Italian instruction-following text generation.
  • Practitioners adapting the model to a downstream Italian task.

Documented uses

  • Try a conversational demo through the project’s Gradio notebook.
  • Load the model in a Python workflow using Transformers and PEFT.
  • Reproduce the project’s training workflow or fine-tune on a task-specific Italian dataset.
Before you choose

Requirements & limits

  • The repository provides a Colab-linked Gradio demo and a separate Python loading example.
  • Fine-tuning data must be JSON with instruction, input and output fields.
  • The project warns about potentially harmful outputs and disallows production usage.
  • The repository lists an Apache-2.0 license.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Fine-tune Camoscio on an Italian task

Prepare a task dataset, merge the model checkpoints and run the repository’s fine-tuning script.

Before you start

  • Camoscio and LLaMA checkpoint files
  • A JSON dataset with instruction, input and output fields
  • Repository dependencies installed with pip install -r requirements.txt

Steps

  1. Prepare the task data as a JSON file using the documented instruction, input and output format.

  2. Merge the original LLaMA weights with the Camoscio LoRA weights by running python scripts/checkpoint_exporter.py.

  3. Run scripts/finetune.py with the merged checkpoint, tokenizer, dataset path and training options.

Expected outcome

The repository describes fine-tuning Camoscio on a downstream Italian task and writing outputs to a chosen output directory.

Guide limitations

  • The repository marks the model for research use and states that production usage is not allowed.
  • The project warns that generated output may be biased, offensive or otherwise harmful.
  • The README’s example command includes placeholder paths that must be replaced.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
17 October 2026
Original website title
teelinsan/camoscio · Official homepage title · retained in its source language
How we review sourcesSuggest a correction