#Bpe-tokenizer
Showing 2 of 2 repositories tagged #bpe-tokenizer, ranked by stars
sefineh-ai
Amharic-Tokenizer
Syllable-aware BPE tokenizer for the Amharic language (አማርኛ) – fast, accurate, trainable.
Score
0
★ 100
⑂ 15
—
Python
JustVugg
nanoeuler
GPT-2-style LLM built from scratch in C/CUDA with hand-written backprop, BPE tokenizer, FlashAttention, pretraining, and SFT.
Score
0
★ 79
⑂ 9
+1/day
Cuda