I need some help to implement an end-to-end retrieval training feature for the rag with Ray. How can I run document encoding and indexing with an updated doc-encoder (context encoder network that kept frozen in the original RAG) using a Ray actor separated from the main training process? How can I…

End to end training of RAG retriever with RAY

Shamane_Siriwardhana February 25, 2021, 11:24pm 7

Hey @amogkam yes I have implemented. But would love to get a review, because I really do not know whether it works since I haven’t tested this with very large-scale training.

Topic		Replies	Views
`map_batches` fails with Huggingface NER pipeline Ray Data	0	39	November 19, 2024
Launching distributed training from within an actor	4	293	August 30, 2023
Unstable actors on GPU	4	138	October 10, 2024
Using ray with transformers pipeline for inference Ray Core	0	501	August 19, 2021
Object Storage Management with Ray Actor Tasks that does not need to be saved once executed Ray Core	4	234	February 28, 2024

End to end training of RAG retriever with RAY

Related topics