Skip to content
View as Markdown

tinker_cookbook.tokenizer_utils

Generated from tinker-cookbook at commit 8486b8d. Source links point at that snapshot.

Class Description
TmlTokenizer Structural type for the underlying tml_renderers tokenizer.
SupportsEncodeOrdinary
SupportsTmlTokenizer Tokenizer facade that exposes an underlying TmlTokenizer.
TmlRenderersTokenizerAdapter Small tokenizer facade for tml_renderers models used through cookbook.
Function Description
ensure_tml_renderers_importable() Make tml_renderers importable, falling back to TML_RENDERERS_SOURCE_DIR.
encode_without_special_tokens() Encode raw text without adding model-level BOS/EOS tokens.
is_tml_renderers_model() Return whether tml-renderers tokenizes model_name.
register_tokenizer() Register a custom tokenizer factory.
get_registered_tokenizer_names() Return a list of all registered custom tokenizer names.
is_tokenizer_registered() Check if a tokenizer name is registered.
unregister_tokenizer() Unregister a custom tokenizer factory.
get_tokenizer() Get a tokenizer by name.