requirements.txt

# GRACE training/eval virtual environment — pinned package versions
#
# Captured from: `pip list` on the qwen3vl venv used to produce the released
# FoeverBLUE/Qwen3-VL-2B-GRACE-{BF16,W8G128,W4G128} checkpoints.
#
# System assumptions (Leonardo / CINECA, A100-80GB):
#   - Python 3.11
#   - CUDA 12.3 + GCC 12.2 (host modules; see scripts/*.slurm `module load`)
#   - PyTorch 2.5.1 built against CUDA 12.1 runtime wheels
#     (the cu121 wheels are forward-compatible with the host CUDA 12.3 driver)
#   - flash-attn 2.7.2.post1 (requires CUDA toolkit available at install time)
#   - DeepSpeed 0.15.4 (ZeRO-2 used for all reference runs)
#
# Recommended install order (flash-attn must be built AFTER torch):
#   pip install torch==2.5.1 torchvision==0.20.1 --index-url https://download.pytorch.org/whl/cu121
#   pip install -r requirements.txt
#   pip install flash-attn==2.7.2.post1 --no-build-isolation
#   pip install -e qwen-vl-utils/

accelerate==1.13.0
aiohappyeyeballs==2.6.2
aiohttp==3.13.5
aiosignal==1.4.0
annotated-doc==0.0.4
annotated-types==0.7.0
antlr4-python3-runtime==4.7.2
anyio==4.13.0
attrs==26.1.0
av==15.1.0
black==26.5.1
certifi==2026.5.20
cfgv==3.5.0
chardet==6.0.0.post1
charset-normalizer==3.4.7
click==8.4.0
colorama==0.4.6
DataProperty==1.1.1
datasets==4.8.5
deepspeed==0.15.4
dill==0.4.1
distlib==0.4.0
distro==1.9.0
einops==0.8.2
et_xmlfile==2.0.0
evaluate==0.4.6
filelock==3.29.0
flash-attn==2.7.2.post1
frozenlist==1.8.0
fsspec==2026.2.0
ftfy==6.3.1
gitdb==4.0.12
GitPython==3.1.50
h11==0.16.0
hf_transfer==0.1.9
hf-xet==1.5.0
hjson==3.1.0
httpcore==1.0.9
httpx==0.28.1
huggingface_hub==1.15.0
identify==2.6.19
idna==3.15
isort==8.0.1
Jinja2==3.1.6
jiter==0.15.0
joblib==1.5.3
jsonlines==4.0.0
latex2sympy2==1.9.1
lmms_eval==0.7.1
loguru==0.7.3
lxml==6.1.1
markdown-it-py==4.2.0
MarkupSafe==3.0.3
math-verify==0.1.0
mbstrdecoder==1.1.5
mdurl==0.1.2
mpmath==1.3.0
msgpack==1.1.2
multidict==6.7.1
multiprocess==0.70.19
mypy_extensions==1.1.0
networkx==3.6.1
ninja==1.13.0
nltk==3.9.4
nodeenv==1.10.0
numexpr==2.14.1
numpy==2.4.4
nvidia-cublas-cu12==12.1.3.1
nvidia-cuda-cupti-cu12==12.1.105
nvidia-cuda-nvrtc-cu12==12.1.105
nvidia-cuda-runtime-cu12==12.1.105
nvidia-cudnn-cu12==9.1.0.70
nvidia-cufft-cu12==11.0.2.54
nvidia-curand-cu12==10.3.2.106
nvidia-cusolver-cu12==11.4.5.107
nvidia-cusparse-cu12==12.1.0.106
nvidia-nccl-cu12==2.21.5
nvidia-nvjitlink-cu12==12.9.86
nvidia-nvtx-cu12==12.1.105
openai==2.38.0
opencv-python-headless==4.13.0.92
openpyxl==3.1.5
packaging==26.2
pandas==3.0.3
pathspec==1.1.1
pathvalidate==3.3.1
peft==0.19.1
pillow==12.2.0
platformdirs==4.9.6
portalocker==3.2.0
pre_commit==4.6.0
propcache==0.5.2
protobuf==6.33.6
psutil==7.2.2
py-cpuinfo==9.0.0
pyarrow==24.0.0
pybind11==3.0.4
pycocoevalcap==1.2
pycocotools==2.0.11
pydantic==2.13.4
pydantic_core==2.46.4
Pygments==2.20.0
pytablewriter==1.2.1
python-dateutil==2.9.0.post0
python-discovery==1.3.1
python-dotenv==1.2.2
pytokens==0.4.1
pytz==2026.2
PyYAML==6.0.3
regex==2026.5.9
requests==2.34.2
rich==15.0.0
sacrebleu==2.6.0
safetensors==0.7.0
scikit-learn==1.8.0
scipy==1.17.1
sentence-transformers==5.5.1
sentencepiece==0.2.1
sentry-sdk==2.60.0
setuptools==82.0.1
shellingham==1.5.4
six==1.17.0
smmap==5.0.3
sniffio==1.3.1
sympy==1.13.1
tabledata==1.3.5
tabulate==0.10.0
tcolorpy==0.1.7
tenacity==9.1.4
threadpoolctl==3.6.0
tiktoken==0.13.0
timm==1.0.27
tokenizers==0.22.2
torch==2.5.1
torchvision==0.20.1
tqdm==4.67.3
tqdm-multiprocess==0.0.11
transformers==5.9.0
transformers-stream-generator==0.0.5
triton==3.1.0
typepy==1.3.5
typer==0.25.1
typing_extensions==4.15.0
typing-inspection==0.4.2
urllib3==2.7.0
virtualenv==21.3.3
wandb==0.25.0
wcwidth==0.7.0
wheel==0.47.0
xxhash==3.7.0
yarl==1.24.2
yt-dlp==2026.3.17
zss==1.2.0
zstandard==0.25.0