{ lib, buildPythonPackage, fetchFromGitHub, pythonOlder, setuptools, # propagated build inputs filelock, huggingface-hub, numpy, protobuf, packaging, pyyaml, regex, requests, tokenizers, safetensors, tqdm, # optional dependencies diffusers, scikit-learn, tensorflow, onnxconverter-common, opencv4, tf2onnx, torch, accelerate, faiss, datasets, jax, jaxlib, flax, optax, ftfy, onnxruntime, onnxruntime-tools, cookiecutter, sagemaker, fairscale, optuna, ray, pydantic, uvicorn, fastapi, starlette, librosa, phonemizer, torchaudio, pillow, timm, torchvision, av, sentencepiece, }: buildPythonPackage rec { pname = "transformers"; version = "4.41.2"; pyproject = true; disabled = pythonOlder "3.8"; src = fetchFromGitHub { owner = "huggingface"; repo = "transformers"; rev = "refs/tags/v${version}"; hash = "sha256-Y3WYO+63n/ATT9RmRJ9JLuhFWUiuBO4NL2KzvzELi+M="; }; build-system = [ setuptools ]; dependencies = [ filelock huggingface-hub numpy packaging pyyaml regex requests tokenizers safetensors tqdm ]; passthru.optional-dependencies = let audio = [ librosa # pyctcdecode phonemizer # kenlm ]; vision = [ pillow ]; in { agents = [ diffusers accelerate datasets torch sentencepiece opencv4 pillow ]; ja = [ # fugashi # ipadic # rhoknp # sudachidict_core # sudachipy # unidic # unidic_lite ]; sklearn = [ scikit-learn ]; tf = [ tensorflow onnxconverter-common tf2onnx # tensorflow-text # keras-nlp ]; torch = [ torch accelerate ]; retrieval = [ faiss datasets ]; flax = [ jax jaxlib flax optax ]; tokenizers = [ tokenizers ]; ftfy = [ ftfy ]; onnxruntime = [ onnxruntime onnxruntime-tools ]; onnx = [ onnxconverter-common tf2onnx onnxruntime onnxruntime-tools ]; modelcreation = [ cookiecutter ]; sagemaker = [ sagemaker ]; deepspeed = [ # deepspeed accelerate ]; fairscale = [ fairscale ]; optuna = [ optuna ]; ray = [ ray ] ++ ray.optional-dependencies.tune-deps; # sigopt = [ sigopt ]; # integrations = ray ++ optuna ++ sigopt; serving = [ pydantic uvicorn fastapi starlette ]; audio = audio; speech = [ torchaudio ] ++ audio; torch-speech = [ torchaudio ] ++ audio; tf-speech = audio; flax-speech = audio; timm = [ timm ]; torch-vision = [ torchvision ] ++ vision; # natten = [ natten ]; # codecarbon = [ codecarbon ]; video = [ # decord av ]; sentencepiece = [ sentencepiece protobuf ]; }; # Many tests require internet access. doCheck = false; pythonImportsCheck = [ "transformers" ]; meta = { homepage = "https://github.com/huggingface/transformers"; description = "Natural Language Processing for TensorFlow 2.0 and PyTorch"; mainProgram = "transformers-cli"; changelog = "https://github.com/huggingface/transformers/releases/tag/v${version}"; license = lib.licenses.asl20; platforms = lib.platforms.unix; maintainers = with lib.maintainers; [ pashashocky happysalada ]; }; }