RAG Fine-tuned Model: `gemma-3-270m-it-RAG-finetuned-202510190155`

This model was fine-tuned using the CPU RAG Fine-Tuner Space.

Model Details

• Base Model: `google/gemma-3-270m-it`
• Fine-Tuning Method: Retrieval-Augmented Generation (RAG) on a CPU. The model was trained to answer questions based on retrieved context.

Training Data

• Dataset: `openai/gsm8k`
• Dataset Configuration: `main`
• Data Slice: Rows `0` to `500` were used.
• Question Column: `question`
• Answer Column: `answer`

Training Hyperparameters

• Learning Rate: `2e-05`
• Epochs: `1`
• Batch Size: `1`

How to Use

This model expects prompts to be formatted in a specific RAG chat structure. The context should be retrieved from a knowledge base built from the training data.

Prompt Template

Use the following context to answer the question.
Context:
---
[Retrieved Document 1]
---
[Retrieved Document 2]
---
...

Question:
[Your Question]

Inference

You can use this model directly in a new RAG Inference Space by simply pasting the model repository ID: `broadfield-dev/gemma-3-270m-it-RAG-finetuned-202510190155`.