PyPI page
Home page
License:
MIT License
Copyright (c) 2026 LightSeek Foundation <contact@lightseek.org>
Permission is hereby granted, free of charge, to any person obtaining a copy
of th...
Summary:
TokenSpeed is a speed-of-light LLM inference engine.
Latest version:
0.1.1
Required dependencies:
aiohttp
|
compressed-tensors
|
dill
|
einops
|
fastapi
|
grpcio-health-checking
|
grpcio-reflection
|
hf-transfer
|
huggingface-hub
|
modelscope
|
msgspec
|
ninja
|
numpy
|
openai
|
openai-harmony
|
orjson
|
packaging
|
partial-json-parser
|
peft
|
pillow
|
prometheus-client
|
psutil
|
py-spy
|
pybase64
|
pybind11
|
pydantic
|
pytest-asyncio
|
python-multipart
|
pyzmq
|
requests
|
setproctitle
|
tiktoken
|
tokenspeed-kernel
|
tokenspeed-mooncake
|
tokenspeed-scheduler
|
tokenspeed-smg
|
tokenspeed-smg-grpc-proto
|
tokenspeed-smg-grpc-servicer
|
torch
|
torch-memory-saver
|
torchvision
|
tqdm
|
transformers
|
uv
|
uvicorn
|
uvloop
|
viztracer
|
xgrammar
Optional dependencies:
instanttensor
Downloads last day:
10
Downloads last week:
59
Downloads last month:
136
7-day smoothing