mt5-base-trimmed-ja / README.md
asahi417's picture
commit files to HF hub
fd70b64

Vocabulary Trimmed google/mt5-base: vocabtrimmer/mt5-base-trimmed-ja

This model is a trimmed version of google/mt5-base by vocabtrimmer, a tool for trimming vocabulary of language models to compress the model size. Following table shows a summary of the trimming process.

google/mt5-base vocabtrimmer/mt5-base-trimmed-ja
parameter_size_full 582,401,280 391,616,256
parameter_size_embedding 384,172,032 193,387,008
vocab_size 250,112 125,903
compression_rate_full 100.0 67.24
compression_rate_embedding 100.0 50.34

Following table shows the parameter used to trim vocabulary.

language dataset dataset_column dataset_name dataset_split target_vocab_size min_frequency
ja vocabtrimmer/mc4_validation text ja validation 2