Llama 3.2 Latin

🚀 Model Card for Llama-3.2-Latin

hathibelagal/llama-3.2-latin is a fine - tuned LLaMA - 3.2 - 3B model optimized for generating and understanding Latin text from ancient to modern Neo - Latin. It can accurately handle tenses, cases, and complex sentence structures.

🚀 Quick Start

hathibelagal/llama-3.2-latin is a finetuned version of the LLaMA-3.2-3B model. It is designed to generate and understand Latin text across different historical periods, from ancient to modern Neo - Latin. Currently, it can generate content with accurate use of tenses (such as pluperfect and subjunctive), cases, and complex structures (such as concessive and temporal clauses).

✨ Features

Intended Use

• Suitable Tasks: This model is suitable for tasks related to Latin text generation, translation, or analysis. For example, it can generate Classical Latin prose, complete sentences, or assist in Latin education. Due to coherence limitations, it works best with short, context - specific prompts.
• Not Recommended: It is not recommended for long - form narrative generation or tasks that demand strict contextual consistency until its coherence is improved.

Ethical Considerations

• Bias: The model may reflect biases present in the training dataset. For instance, an overrepresentation of certain Latin styles (e.g., ecclesiastical Latin) can lead to tonal shifts in the output.
• Usage: When using the generated text, especially in educational or scholarly contexts, it should be carefully reviewed for accuracy.

💻 Usage Examples

Basic Usage

from transformers import AutoTokenizer, AutoModelForCausalLM
import torch

repo_id = "hathibelagal/llama-3.2-latin"

tokenizer = AutoTokenizer.from_pretrained(repo_id)
model = AutoModelForCausalLM.from_pretrained(
    repo_id,
    torch_dtype=torch.float16,
    device_map="auto"
)

prompt = "Libellus vere aureus"
inputs = tokenizer.encode(prompt, return_tensors="pt")
inputs = inputs.to(model.device)

outputs = model.generate(
    inputs,
    max_new_tokens=100,
    do_sample=True,
    top_p=0.9,
    temperature=0.6,
    repetition_penalty=1.2
)

print(tokenizer.batch_decode(
    outputs, skip_special_tokens=True)[0])

📚 Documentation

Model Information