also shout out to @clifapt for running BM25 via ES on LoCo first: https://
x.com/clifapt/status
/1755303087873429667
… my implementation seems to be slightly less performant than ElasticSearch; not sure why. maybe because i'm using a subword tokenizer, or suboptimal hyperparams?
@jxmnop
-
BM25 Search Implementation Performance Comparison with ElasticSearch
By
–
-
BM25_PT: Matrix Multiplication for Efficient Search Scoring
By
–
my library is here, bm25_pt: http://
github.com/jxmorris12/bm2
5_pt
… the key insight is that you can reduce the BM25 scoring to a single matrix multiplication. everything in the big fraction on the right side here can be stored in a big matrix of scores then, bm25(q) = bag(q) @ scores.T -
Fast GPU-Enabled BM25 Implementation in PyTorch Achieves SOTA
By
–
implemented a fast, GPU-enabled BM25 in pytorch! BM25 is a simple search algorithm from the 70s that works as well as neural networks for most search problems; for all the advances we've made in neural text retrieval, it's still around got near SOTA on stanford LoCO benchmark
-
GTR Model Embeddings Bug Fix and Post-Processing Correction
By
–
unfortunately that's due to a bug I made when training the model; the gtr embeddings i use are missing a last post-processing step. someone trained a fixed model, more info here https://
github.com/jxmorris12/vec
2text/issues/28
… -
Embeddings Distribution and Decompression in AI Models
By
–
I also wonder if taking the mean of several embeddings makes them “out of distribution” somehow im really not sure, about the muster question! I convinced myself it’s because embeddings are sort of “compressed” and it takes a lot of effort to decompress them
-
2024 Will Likely Bring AI Advancement Progress
By
–
i think extrapolating that 2024 will bring at least one advancement in AI is a pretty reasonable bet
-
Bulgarian Developer Makes LLMs Run on CPU with llama.cpp
By
–
2022:
– oh nooo!!! you can't run language models on cpu! you need an expensive nvidia GPU and special CUDA kernels and–
– *one bulgarian alpha chad sits down and writes some c++ code to run LLMs on cpu* – code works fine (don't need a GPU), becomes llama.cpp 2023:
– oh noo!! -
Optimizing Reconstruction Steps for Better AI Model Performance
By
–
could you try using fewer reconstruction steps? that much actually work better in this case