Maître-Granite-3.2-2b-instruct

This model is a basic fine-tuned bilingual English-French version of ibm-granite/granite-3.2-2b-instruct. It has been trained on the agentlans/en-fr-debut-kit dataset. This model does well for its size and its small amount of training data.

Ce modèle est une version bilingue anglais-français de base et optimisée du modèle ibm-granite/granite-3.2-2b-instruct. Il a été entraîné sur le jeu de données agentlans/en-fr-debut-kit. Ce modèle est performant compte tenu de sa taille et de sa faible quantité de données d'entraînement.

If you want a more complete French chatbot, please consider these models:

Si vous souhaitez un chatbot français plus complet, veuillez considérer ces modèles :

• croissantllm/CroissantLLMChat-v0.1
• OpenLLM-France/Lucie-7B-Instruct-v1.1

Model Details

• Base Model: ibm-granite/granite-3.2-2b-instruct
• Fine-tuned from: ibm-granite/granite-3.2-2b-instruct
• Model Type: Transformer-based language model
• Languages: English (en), French (fr)
• License: Apache 2.0

Intended Use

This model is designed for:

• Instruction following in both English and French.
• Machine translation between English and French.
• General-purpose language tasks in English and French.

Training Data

The model was fine-tuned on the agentlans/en-fr-debut-kit dataset. This dataset is specifically designed to facilitate bilingual English-French language tasks. More details about the dataset can be found on its Hugging Face Dataset page.

Training Procedure

The training process included the following stages:

1. Pretraining
2. Supervised Fine-tuning (SFT)
3. Direct Preference Optimization (DPO)

Further details about the specific hyperparameters, training infrastructure, and optimization techniques will be added later.