RAG (Retrieval-Augmented Generation) has three types:
1. Frozen Model RAG: These models are widely seen across the industry and are primarily used for proof of concept (POC).
2. Semi-Frozen Model RAG: Intelligent retrievers are applied, attempting to adapt them in some way. The LLM is not modified; only the retrievers are operated and combined with the final output.
3. Fully Trainable RAG: End-to-end training is quite challenging, but if done correctly, it can provide optimal performance. However, it is certainly very resource-intensive.
RAG (Retrieval-Augmented Generation) has three types:
1. Frozen Model RAG: These models are widely seen across the industry and are primarily used for proof of concept (POC).
2. Semi-Frozen Model RAG: Intelligent retrievers are applied, attempting to adapt them in some way. The LLM is not modified; only the retrievers are operated and combined with the final output.
3. Fully Trainable RAG: End-to-end training is quite challenging, but if done correctly, it can provide optimal performance. However, it is certainly very resource-intensive.