Perf/bpe merge - #2241
Open
ArthurZucker wants to merge 84 commits into
Open
Conversation
|
The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update. |
ArthurZucker
changed the base branch from
feat/bpe-cache-min
to
feat/train_encode_split
August 3, 2026 08:09
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
PipelineTokenizer benchmark
10 / 10 models supported — PipelineTokenizer vs
tokenizersv0.23.1 (latest release) · ~10 kB inputs · add_special_tokens on · single thread + 1/2/4/8/max-thread sweepa42edc914 · 2026-08-03 10:02 UTC· Intel(R) Xeon(R) Platinum 8375C CPU @ 2.90GHz · 48 coresvs base branch (
d183afea5) — per-model geomean ×speedup of this PR's PipelineTokenizer against the base branch's; regressions in red.Decode
Round-trip: v0.23.1
encode_fastproduces the id streams (same fixtures,add_special_tokens=true); both implementations decode those SAME ids withskip_special_tokens=false. MB/s counts decoded text bytes.bert-base-uncased — normalizer-heavy WordPiece · ×4.88 vs v0.23.1 · ×0.99 vs base · decode pending
Memory (RSS MB, load+encode): v0.23.1 12+0 (peak 12) · Pipeline 8+0 (peak 17)
deepseek-v4 — deepseek 3-regex split-heavy byte-level BPE · ×6.98 vs v0.23.1 · ×1.70 vs base · decode pending
Memory (RSS MB, load+encode): v0.23.1 62+0 (peak 67) · Pipeline 92+0 (peak 91)
Pre-tokenize:
classify + fsmvs regex engines — ns/byte, lower better. The fsm is the scalar jump-table in both pipe columns; SIMD / scalar is the classify pass (regex pre-tokenizers have no SIMD fsm).×vs= engine ÷ our pipeline (SIMD / scalar classify);onig&pcre2(JIT) are C,fancyis pure-Rust fancy-regex,logosis a compile-time DFA lexer (approximate grammar; n/a for deepseek).gemma-4 — byte-fallback BPE, Metaspace-style split (gemma-4) · ×2.94 vs v0.23.1 · ×1.81 vs base · decode pending
Memory (RSS MB, load+encode): v0.23.1 304+0 (peak 370) · Pipeline 274+0 (peak 370)
gpt2 — gpt2 ByteLevel regex · ×13.25 vs v0.23.1 · ×1.66 vs base · decode pending
Memory (RSS MB, load+encode): v0.23.1 25+2 (peak 27) · Pipeline 32+0 (peak 31)
Pre-tokenize:
classify + fsmvs regex engines — ns/byte, lower better. The fsm is the scalar jump-table in both pipe columns; SIMD / scalar is the classify pass (regex pre-tokenizers have no SIMD fsm).×vs= engine ÷ our pipeline (SIMD / scalar classify);onig&pcre2(JIT) are C,fancyis pure-Rust fancy-regex,logosis a compile-time DFA lexer (approximate grammar; n/a for deepseek).gpt-oss — o200k-regex byte-level BPE (gpt-oss) · ×7.50 vs v0.23.1 · ×1.55 vs base · decode pending
Memory (RSS MB, load+encode): v0.23.1 241+0 (peak 315) · Pipeline 234+0 (peak 316)
Pre-tokenize:
classify + fsmvs regex engines — ns/byte, lower better. The fsm is the scalar jump-table in both pipe columns; SIMD / scalar is the classify pass (regex pre-tokenizers have no SIMD fsm).×vs= engine ÷ our pipeline (SIMD / scalar classify);onig&pcre2(JIT) are C,fancyis pure-Rust fancy-regex,logosis a compile-time DFA lexer (approximate grammar; n/a for deepseek).glm-5.2 — cl100k-variant regex byte-level BPE (glm-5.2) · ×9.00 vs v0.23.1 · ×1.46 vs base · decode pending
Memory (RSS MB, load+encode): v0.23.1 169+0 (peak 230) · Pipeline 170+0 (peak 231)
Pre-tokenize:
classify + fsmvs regex engines — ns/byte, lower better. The fsm is the scalar jump-table in both pipe columns; SIMD / scalar is the classify pass (regex pre-tokenizers have no SIMD fsm).×vs= engine ÷ our pipeline (SIMD / scalar classify);onig&pcre2(JIT) are C,fancyis pure-Rust fancy-regex,logosis a compile-time DFA lexer (approximate grammar; n/a for deepseek).llama-2 — model-bounded BPE, no pre-tokenizer · ×5.88 vs v0.23.1 · ×1.62 vs base · decode pending
Memory (RSS MB, load+encode): v0.23.1 19+0 (peak 23) · Pipeline 29+0 (peak 29)
llama-3 — cl100k-regex byte-level BPE (llama-3), single regex · ×9.70 vs v0.23.1 · ×1.49 vs base · decode pending
Memory (RSS MB, load+encode): v0.23.1 73+0 (peak 95) · Pipeline 108+0 (peak 108)
Pre-tokenize:
classify + fsmvs regex engines — ns/byte, lower better. The fsm is the scalar jump-table in both pipe columns; SIMD / scalar is the classify pass (regex pre-tokenizers have no SIMD fsm).×vs= engine ÷ our pipeline (SIMD / scalar classify);onig&pcre2(JIT) are C,fancyis pure-Rust fancy-regex,logosis a compile-time DFA lexer (approximate grammar; n/a for deepseek).mistral-small-4 — tekken byte-level BPE, 1k added specials (mistral-small-4) · ×6.92 vs v0.23.1 · ×1.34 vs base · decode pending
Memory (RSS MB, load+encode): v0.23.1 152+0 (peak 194) · Pipeline 145+0 (peak 194)
Pre-tokenize:
classify + fsmvs regex engines — ns/byte, lower better. The fsm is the scalar jump-table in both pipe columns; SIMD / scalar is the classify pass (regex pre-tokenizers have no SIMD fsm).×vs= engine ÷ our pipeline (SIMD / scalar classify);onig&pcre2(JIT) are C,fancyis pure-Rust fancy-regex,logosis a compile-time DFA lexer (approximate grammar; n/a for deepseek).t5-base — Unigram + Metaspace · ×2.58 vs v0.23.1 · ×1.03 vs base · decode pending
Memory (RSS MB, load+encode): v0.23.1 34+2 (peak 36) · Pipeline 61+1 (peak 66)