98 lines
3.3 KiB
Python
98 lines
3.3 KiB
Python
from __future__ import annotations
|
|
|
|
from .core import Encoding
|
|
from .registry import get_encoding
|
|
|
|
# TODO: these will likely be replaced by an API endpoint
|
|
MODEL_PREFIX_TO_ENCODING: dict[str, str] = {
|
|
# chat
|
|
"gpt-4-": "cl100k_base", # e.g., gpt-4-0314, etc., plus gpt-4-32k
|
|
"gpt-3.5-turbo-": "cl100k_base", # e.g, gpt-3.5-turbo-0301, -0401, etc.
|
|
"gpt-35-turbo-": "cl100k_base", # Azure deployment name
|
|
# fine-tuned
|
|
"ft:gpt-4": "cl100k_base",
|
|
"ft:gpt-3.5-turbo": "cl100k_base",
|
|
"ft:davinci-002": "cl100k_base",
|
|
"ft:babbage-002": "cl100k_base",
|
|
}
|
|
|
|
MODEL_TO_ENCODING: dict[str, str] = {
|
|
# chat
|
|
"gpt-4": "cl100k_base",
|
|
"gpt-3.5-turbo": "cl100k_base",
|
|
"gpt-35-turbo": "cl100k_base", # Azure deployment name
|
|
# base
|
|
"davinci-002": "cl100k_base",
|
|
"babbage-002": "cl100k_base",
|
|
# embeddings
|
|
"text-embedding-ada-002": "cl100k_base",
|
|
# DEPRECATED MODELS
|
|
# text (DEPRECATED)
|
|
"text-davinci-003": "p50k_base",
|
|
"text-davinci-002": "p50k_base",
|
|
"text-davinci-001": "r50k_base",
|
|
"text-curie-001": "r50k_base",
|
|
"text-babbage-001": "r50k_base",
|
|
"text-ada-001": "r50k_base",
|
|
"davinci": "r50k_base",
|
|
"curie": "r50k_base",
|
|
"babbage": "r50k_base",
|
|
"ada": "r50k_base",
|
|
# code (DEPRECATED)
|
|
"code-davinci-002": "p50k_base",
|
|
"code-davinci-001": "p50k_base",
|
|
"code-cushman-002": "p50k_base",
|
|
"code-cushman-001": "p50k_base",
|
|
"davinci-codex": "p50k_base",
|
|
"cushman-codex": "p50k_base",
|
|
# edit (DEPRECATED)
|
|
"text-davinci-edit-001": "p50k_edit",
|
|
"code-davinci-edit-001": "p50k_edit",
|
|
# old embeddings (DEPRECATED)
|
|
"text-similarity-davinci-001": "r50k_base",
|
|
"text-similarity-curie-001": "r50k_base",
|
|
"text-similarity-babbage-001": "r50k_base",
|
|
"text-similarity-ada-001": "r50k_base",
|
|
"text-search-davinci-doc-001": "r50k_base",
|
|
"text-search-curie-doc-001": "r50k_base",
|
|
"text-search-babbage-doc-001": "r50k_base",
|
|
"text-search-ada-doc-001": "r50k_base",
|
|
"code-search-babbage-code-001": "r50k_base",
|
|
"code-search-ada-code-001": "r50k_base",
|
|
# open source
|
|
"gpt2": "gpt2",
|
|
}
|
|
|
|
|
|
def encoding_name_for_model(model_name: str) -> str:
|
|
"""Returns the name of the encoding used by a model.
|
|
|
|
Raises a KeyError if the model name is not recognised.
|
|
"""
|
|
encoding_name = None
|
|
if model_name in MODEL_TO_ENCODING:
|
|
encoding_name = MODEL_TO_ENCODING[model_name]
|
|
else:
|
|
# Check if the model matches a known prefix
|
|
# Prefix matching avoids needing library updates for every model version release
|
|
# Note that this can match on non-existent models (e.g., gpt-3.5-turbo-FAKE)
|
|
for model_prefix, model_encoding_name in MODEL_PREFIX_TO_ENCODING.items():
|
|
if model_name.startswith(model_prefix):
|
|
return model_encoding_name
|
|
|
|
if encoding_name is None:
|
|
raise KeyError(
|
|
f"Could not automatically map {model_name} to a tokeniser. "
|
|
"Please use `tiktoken.get_encoding` to explicitly get the tokeniser you expect."
|
|
) from None
|
|
|
|
return encoding_name
|
|
|
|
|
|
def encoding_for_model(model_name: str) -> Encoding:
|
|
"""Returns the encoding used by a model.
|
|
|
|
Raises a KeyError if the model name is not recognised.
|
|
"""
|
|
return get_encoding(encoding_name_for_model(model_name))
|