vision/smolvlm2/pyproject.toml

[project]
name = "smolvlm"
version = "0.1.0"
description = "Fine-tuning smolvlm models with multi-modal support (images + videos)."
readme = "README.md"
license = { file = "LICENSE" }
authors = [
  { name = "Orr Zohar", email = "orrzohar@stanford.edu" }
]
keywords = ["NLP", "deep-learning", "transformers", "multi-modal", "vision-language", "video"]
dependencies = [
  "torch>=2.1.2",
  "torchvision",
  "transformers>=4.47.1",
  "datasets>=2.0.0",
  "peft>=0.2.0",
  "deepspeed>=0.13.5",
  "safetensors>=0.2.0",
  "bitsandbytes>=0.37.0",
  "accelerate>=0.33.0",
  "ujson>=5.0.0",
  "numpy>=1.21.0",
  "opencv-python",
  "chardet",
  "decord",
  "liger-kernel",
  "tabulate",
  "num2words",
]

[project.optional-dependencies]
wandb = [
  "wandb>=0.12.0",
]
tensorboard = [
  "tensorboard>=2.9.0",
]
train = ["ninja", "wandb", "ipdb"]

[build-system]
requires = ["setuptools>=61.0", "wheel"]
build-backend = "setuptools.build_meta"

[tool.setuptools.packages.find]
where = ["."]
include = ["smolvlm*", "scripts*"]