Minghan Li

icon indicating a website link alexlimh.github.io icon indicating an email [email protected]

icon company University of Waterloo icon location Waterloo, ON icon location Information Retrieval Open-Domain Question Answering Retrieval-Augmented LLMs

Results 5 issues of Minghan Li

Batching for IndexReader's methods

Does anyone know if there's a batching method for IndexReader.compute_query_document_score in anserini? The original method seems to run very slow, roughly 6 hrs for ~500k query-doc pairs on an Intel(R)...

Build inverted indexes on the fly

4 comment

Currently pyserini only supports building indexes on the document collections stored in the disk, but it would be nice to also support building indexes on the fly, i.e., using GPT-2...

MLM head pre-trained weights of character-bert

Hi, I just found that the input embeddings of character-bert encoder are not tied with the output MLM weights. I would appreciate it if you could also release the pre-trained...

Correction for a mistake in 'last2avg'

Dear BERT-Flow authors, I notice there's a mistake in your code where you attempt to average the last 2 layers of BERT but instead you average the first and last...

Shared Weights in ConCaps Layer

Nice implementation of the capsule network with EM routing! However, I do have a question about the function mat_transform() in the ConvCaps layer. In your implementation the weights are not...