Tip: Alternative installation method using uv

[kawamou]

Summary

I successfully set up this project using uv instead of conda by modifying the existing pyproject.toml file. Sharing this for anyone who prefers uv over conda.

Environment

• OS: Windows WSL (Ubuntu 24.04.1 LTS)
• Python: 3.11+

What I did

I modified the existing pyproject.toml file to include all necessary dependencies. This allows users to install the project using uv commands as an alternative to the conda installation method described in the README.

Installation commands with uv

uv sync --no-dev
uv sync --dev --no-build-isolation

How to set it up

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

[project]
name = "starvector"
version = "1.0"
description = "Generating Scalable Vector Graphics Code from Images and Text"
readme = "README.md"
requires-python = ">=3.11"
classifiers = [
    "Programming Language :: Python :: 3",
    "License :: OSI Approved :: Apache Software License",
]
dependencies = [
    "torch==2.5.1",
    "torchvision==0.20.1",
    "transformers==4.49.0",
    "tokenizers==0.21.1",
    "sentencepiece==0.2.0",
    "accelerate",
    "pydantic==2.10",
    "markdown2[all]",
    "numpy<2.0.0",
    "scikit-learn==1.2.2",
    "gradio==3.36.1",
    "gradio_client==0.2.9",
    "requests",
    "httpx==0.24.0",
    "uvicorn",
    "fastapi",
    "svgpathtools==1.6.1",
    "seaborn==0.12.2",
    "taming-transformers",
    "lpips",
    "cairosvg",
    "beautifulsoup4",
    "webcolors",
    "tqdm",
    "omegaconf",
    "open-clip-torch",
    "noise",
    "datasets",
    "scikit-image",
    "fairscale",
    "lxml",
    "torch-fidelity",
    "clip-openai",
    "scipy==1.11.1",
    "sentence-transformers",
    "reportlab",
    "svglib",
    "Pillow",
    "protobuf",
    "openai",
    "hatchling>=1.27.0",
    "editables>=0.5",
    "wheel>=0.45.1",
    "deepspeed>=0.17.1",
]

[project.optional-dependencies]
train = ["deepspeed", "ninja", "wandb"]

[project.urls]
"Homepage" = "https://starvector.github.io"
"Bug Tracker" = "https://github.com/joanrod/starvector/issues"

[tool.setuptools.packages.find]
exclude = ["assets*", "benchmark*", "docs", "dist*", "playground*", "scripts*", "tests*"]

[tool.wheel]
exclude = ["assets*", "benchmark*", "docs", "dist*", "playground*", "scripts*", "tests*"]

[dependency-groups]
dev = [
    "flash-attn==2.7.3",
]

I tested this setup and it works well on my system. I was able to successfully run the example code from the README:

from PIL import Image
from transformers import AutoModelForCausalLM, AutoTokenizer, AutoProcessor
from starvector.data.util import process_and_rasterize_svg
import torch

model_name = "starvector/starvector-8b-im2svg"
starvector = AutoModelForCausalLM.from_pretrained(model_name, torch_dtype=torch.float16, trust_remote_code=True)
processor = starvector.model.processor
tokenizer = starvector.model.svg_transformer.tokenizer
starvector.cuda()
starvector.eval()

image_pil = Image.open('assets/examples/sample-18.png')
image = processor(image_pil, return_tensors="pt")['pixel_values'].cuda()
if not image.shape[0] == 1:
    image = image.squeeze(0)
batch = {"image": image}
raw_svg = starvector.generate_im2svg(batch, max_length=4000)[0]
svg, raster_image = process_and_rasterize_svg(raw_svg)

[HighLiuk]

I get this by doing uv sync --no-dev (MacOS Tahoe 26.0.1)

Using CPython 3.13.3
Creating virtual environment at: .venv
  × Failed to build `flash-attn==2.7.3`
  ├─▶ The build backend returned an error
  ╰─▶ Call to `setuptools.build_meta:__legacy__.build_wheel` failed (exit status: 1)

      [stderr]
      /Users/luca/.cache/uv/builds-v0/.tmpM8mFJu/lib/python3.13/site-packages/setuptools/_vendor/wheel/bdist_wheel.py:4:
      FutureWarning: The 'wheel' package is no longer the canonical location of the 'bdist_wheel' command, and will be
      removed in a future release. Please update to setuptools v70.1 or later which contains an integrated version of
      this command.
        warn(
      Traceback (most recent call last):
        File "<string>", line 14, in <module>
          requires = get_requires_for_build({})
        File "/Users/luca/.cache/uv/builds-v0/.tmpM8mFJu/lib/python3.13/site-packages/setuptools/build_meta.py", line
      333, in get_requires_for_build_wheel
          return self._get_build_requires(config_settings, requirements=[])
                 ~~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
        File "/Users/luca/.cache/uv/builds-v0/.tmpM8mFJu/lib/python3.13/site-packages/setuptools/build_meta.py", line
      301, in _get_build_requires
          self.run_setup()
          ~~~~~~~~~~~~~~^^
        File "/Users/luca/.cache/uv/builds-v0/.tmpM8mFJu/lib/python3.13/site-packages/setuptools/build_meta.py", line
      518, in run_setup
          super().run_setup(setup_script=setup_script)
          ~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^
        File "/Users/luca/.cache/uv/builds-v0/.tmpM8mFJu/lib/python3.13/site-packages/setuptools/build_meta.py", line
      317, in run_setup
          exec(code, locals())
          ~~~~^^^^^^^^^^^^^^^^
        File "<string>", line 22, in <module>
      ModuleNotFoundError: No module named 'torch'

      hint: This error likely indicates that `flash-attn@2.7.3` depends on `torch`, but doesn't declare it as a build
      dependency. If `flash-attn` is a first-party package, consider adding `torch` to its `build-system.requires`.
      Otherwise, either add it to your `pyproject.toml` under:

      [tool.uv.extra-build-dependencies]
      flash-attn = ["torch"]

      or `uv pip install torch` into the environment and re-run with `--no-build-isolation`.
  help: `flash-attn` (v2.7.3) was included because `starvector` (v1.0) depends on `flash-attn==2.7.3`