Pere Lluis

icon indicating a website link https://scholar.google.com/citations?user=HeqN6q8AAAAJ&hl=en icon indicating an email [email protected]

icon company Meta icon location Rome icon location NLP enthusiast. Past: research collaborator at ILLC in University of Amsterdam. Current: PhD at Babelscape and La Sapienza University, Rome

Results 16 comments of Pere Lluis

Create dataloader

Added a PR with Dataloader and Dataset. Unfortunately I could not improve performance much, each video preprocessing takes around 3 minutes, since it needs to do operations per frame to...

Minor question about PAD token and EOS token.

Setting the pad token to eos is an issue on our training as well. What I do not get is how Zephyr was trained with such recipe, since Mistral does...

How to fine-tune?

Please check the https://github.com/SapienzaNLP/relik?tab=readme-ov-file#-before-you-start section in the Readme. If you want to fine-tune the reader on top of one of our already trained models instead of deberta, just point to...

Attempted unscale_ but _scale is None after resuming training.

Not for the moment, but I can try relaunching the same training when we are able to update our codebase to v0.5.

Issues with fine-tuning sapienzanlp/relik-reader-deberta-v3-base-aida

(sorry it took this long to reply) Can you provide the error?

Whether can you provide PROCESSED NYT dataset used for entity relation extraction training

Hi there, sorry for the late reply. Which processed data do you mean? For training you need to follow the readme and create it for both the retriever and the...