Update README.md

README.md

@@ -6,6 +6,14 @@ Cassandre-RAG is a fine-tuned llama-3.1-8b model, built for RAG on French admini
 
 The model was trained on a H100, using these parameters:
 
+## Training Data
+
+The model was fine-tuned on a specialized corpus consisting of:
+
+1. Synthetic queries: Generated from chunks of text extracted from French administrative documents.
+2. Retrieved documents: For each synthetic query, relevant documents were retrieved using the BM25 ranking algorithm.
+3. Generated answers: Responses to the synthetic queries were created based on the retrieved documents.
+
 ### Training Hyperparameters
 - Max Steps: 3000
 - Learning Rate: 3e-4