IPAFuzzv1amav5: Fine-tuned LLaMA 3.2 (1B) for IPA Pronunciations

Overview

IPAFuzzv1amav5 is a fine-tuned LLaMA 3.2 (1B) model trained using Unsloth to generate all possible pronunciations in the International Phonetic Alphabet (IPA) for a given name spelled in English script.

Features

  • Generates multiple possible IPA transcriptions for a given name
  • Handles regional and dialectal variations
  • Supports ambiguous spellings
  • Efficient inference with optimized fine-tuning via Unsloth

Model Details

  • Base Model: LLaMA 3.2 (1B)
  • Fine-tuning Framework: Unsloth
  • Training Data: Names with known IPA transcriptions from multiple linguistic sources
  • Objective: Predict all phonetic variations for a given input name

Example Outputs

Name IPA Transcriptions
Alice /ˈælɪs/, /ˈeɪlɪs/, /ˈælɪs/
John /dʒɒn/, /dʒɔn/, /dʒɑn/

License

MIT License

Acknowledgments

  • Meta for LLaMA 3.2
  • Unsloth for efficient model fine-tuning
  • Open-source linguistic datasets

Uploaded model

  • Developed by: Noeai
  • License: apache-2.0
  • Finetuned from model : unsloth/llama-3.2-1b-instruct-bnb-4bit

Downloads last month
10
GGUF
Model size
1B params
Architecture
llama
Hardware compatibility
Log In to view the estimation

8-bit

Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support