Ahmet Erdem
Ahmet Erdem
When I change the base image to my host machine CUDA version, it works.
Makes sense. Would this be a flag to set? I have only one concern. Most BN implementations etc make safe normalization but someone may have a custom model which assumes...
I have thought about it but I think it will be very difficult to implement rolling count and nunique given the limitations from Numba CUDA functions. But we can implement...
Hi, this model can run on a Kaggle kernel: https://www.kaggle.com/wiki/Scripts. And it finishes within an hour.
Can you please share the code for me to be able to reproduce the issue?
Is this implemented? Another need is to be able too have cosine similarity without l2 normalization option on KNN. This way one can approximate BM25 matching. Since long documents suffer...
CPU Version ```python from beir import util, LoggingHandler from beir.retrieval import models from beir.datasets.data_loader import GenericDataLoader from beir.retrieval.evaluation import EvaluateRetrieval import pandas as pd from sparse_dot_topn import sp_matmul, sp_matmul_topn import...