blog
blog copied to clipboard
HuggingFace Tokenizers as Collate Functions Timing 🤗 🤖 | DeepSchool
HuggingFace Tokenizers as Collate Functions Timing 🤗 🤖 | DeepSchool
Timing comparison of tokenizer as collate function and after batching
https://sachinruk.github.io/blog/pytorch/huggingface/2021/11/17/collate-fn-vs-normal.html
Hey, it seems that the utterances integration is broken on your website so i'm directly commenting here.
Thanks for the blogpost. It is an interesting result. Have you since fully switched to collation outside of the dataloader? Or is there something keeping you to the traditional way of passing collate_fn to the dataloader?