pyproject.toml

[project]
name = "compactor-vllm"
description = "Fast KV Cache Compression for LLMs"
version = "0.0.1"
dependencies = [
    # "triton>=3.5.0",
    "transformers",
    # "torch>=2.9.0",
    "safetensors",
    "tqdm",
    "flash-attn",
    "pytest"
]
requires-python = ">= 3.8"
authors = [
  {name = "Vivek Chari", email = "viveknchari@gmail.com"},
]
[project.optional-dependencies]
evaluate = ["rouge", "pandas",  "fuzzywuzzy"]
[tool.ruff]
exclude = [
    "triton_kernels"
]