Configspace_converter data types

[akalino]

Hello,

I'm facing a datatype issue when running jobs from the GraSH CLI. Specifically, when reading in config files, I am given the following error:

line 42, in get_configspace
CSH.UniformFloatHyperparameter(
File "ConfigSpace/hyperparameters/uniform_float.pyx", line 62, in ConfigSpace.hyperparameters.uniform_float.UniformFloatHyperparameter.init
File "ConfigSpace/hyperparameters/uniform_float.pyx", line 124, in ConfigSpace.hyperparameters.uniform_float.UniformFloatHyperparameter.check_default
TypeError: Expected float, got numpy.float64

This comes from the LR bounds, which I have as follows in my config yaml:

parameters:

• bounds:
  • 0.0003
  • 1.0
    log_scale: true
    name: train.optimizer_args.lr
    type: range

My environment is as described below, using two gpus for shared compute:

name: kge
channels:

• conda-forge
  dependencies:
• _libgcc_mutex=0.1=conda_forge
• _openmp_mutex=4.5=2_gnu
• bzip2=1.0.8=hd590300_5
• ca-certificates=2024.2.2=hbcca054_0
• ld_impl_linux-64=2.40=h41732ed_0
• libexpat=2.6.1=h59595ed_0
• libffi=3.4.2=h7f98852_5
• libgcc-ng=13.2.0=h807b86a_5
• libgomp=13.2.0=h807b86a_5
• libnsl=2.0.1=hd590300_0
• libsqlite=3.45.1=h2797004_0
• libuuid=2.38.1=h0b41bf4_0
• libxcrypt=4.4.36=hd590300_1
• libzlib=1.2.13=hd590300_5
• ncurses=6.4=h59595ed_2
• openssl=3.2.1=hd590300_0
• pip=24.0=pyhd8ed1ab_0
• python=3.12.2=hab00c5b_0_cpython
• readline=8.2=h8228510_1
• setuptools=69.1.1=pyhd8ed1ab_0
• tk=8.6.13=noxft_h4845f30_101
• wheel=0.42.0=pyhd8ed1ab_0
• xz=5.2.6=h166bdaf_0
• pip:
  • argparse==1.4.0
  • asttokens==2.4.1
  • ax-platform==0.3.7
  • botorch==0.10.0
  • comm==0.2.1
  • configspace==0.7.1
  • contextlib2==21.6.0
  • decorator==5.1.1
  • execnet==2.0.2
  • executing==2.0.1
  • filelock==3.13.1
  • fsspec==2024.2.0
  • gpytorch==1.11
  • greenlet==3.0.3
  • hpbandster==0.7.4
  • igraph==0.11.4
  • iniconfig==2.0.0
  • ipython==8.22.2
  • ipywidgets==8.1.2
  • jaxtyping==0.2.28
  • jedi==0.19.1
  • jinja2==3.1.3
  • joblib==1.3.2
  • jupyterlab-widgets==3.0.10
  • linear-operator==0.5.1
  • llvmlite==0.42.0
  • markupsafe==2.1.5
  • matplotlib-inline==0.1.6
  • mock==5.1.0
  • more-itertools==10.2.0
  • mpmath==1.3.0
  • multipledispatch==1.0.0
  • mypy-extensions==1.0.0
  • netifaces==0.11.0
  • networkx==3.2.1
  • numba==0.59.0
  • numpy==1.26.4
  • 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==8.9.2.26
  • 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==2.19.3
  • nvidia-nvjitlink-cu12==12.4.99
  • nvidia-nvtx-cu12==12.1.105
  • opt-einsum==3.3.0
  • packaging==24.0
  • pandas==2.2.1
  • parso==0.8.3
  • path==16.10.0
  • path-py==12.5.0
  • patsy==0.5.6
  • pexpect==4.9.0
  • plotly==5.19.0
  • pluggy==1.4.0
  • prompt-toolkit==3.0.43
  • ptyprocess==0.7.0
  • pure-eval==0.2.2
  • pygments==2.17.2
  • pyparsing==3.1.2
  • pyre-extensions==0.0.30
  • pyro-api==0.1.2
  • pyro-ppl==1.9.0
  • pyro4==4.82
  • pytest==8.1.1
  • pytest-shutil==1.7.0
  • python-dateutil==2.9.0.post0
  • python-graphviz==0.20.1
  • pytz==2024.1
  • pyyaml==6.0.1
  • scikit-learn==1.4.1.post1
  • scipy==1.12.0
  • serpent==1.41
  • six==1.16.0
  • sqlalchemy==1.4.52
  • stack-data==0.6.3
  • statsmodels==0.14.1
  • sympy==1.12
  • tenacity==8.2.3
  • termcolor==2.4.0
  • texttable==1.7.0
  • threadpoolctl==3.3.0
  • torch==2.2.1
  • torchviz==0.0.2
  • tqdm==4.66.2
  • traitlets==5.14.1
  • typeguard==2.13.3
  • typing-extensions==4.10.0
  • typing-inspect==0.9.0
  • tzdata==2024.1
  • wcwidth==0.2.13
  • widgetsnbextension==4.0.10
    prefix: /home/alexander-kalinowski/miniforge3/envs/kge

Is there a quick fix I can use to get around this? I have tried to directly cast the relevant hyperparameters to floats with no luck.

[AdrianKs]

Hi,
I am sorry for the very late reply, I just saw your Issue now. I suppose the issue lies in the numpy version.
Around version 1.25 or so they changed something data type related.
I fixed the setup script so that it should install dependencies correctly now:

conda create -n grash python==3.8
conda activate grash
pip install -e .

python -m kge start examples/toy-complex-search-grash.yaml --job.device cpu

I hope this works for you.

It worked for me with the following package versions after installation:

package	version
ax-platform	0.1.19
botorch	0.4.0
ConfigSpace	0.7.1
filelock	3.13.4
fsspec	2024.3.1
gpytorch	1.4.2
graphviz	0.20.3
greenlet	3.0.3
hpbandster	0.7.4
igraph	0.11.4
importlib_metadata	7.1.0
Jinja2	3.1.3
joblib	1.4.0
libkge	0.1
llvmlite	0.33.0
MarkupSafe	2.1.5
mock	5.1.0
more-itertools	10.2.0
mpmath	1.3.0
netifaces	0.11.0
networkx	3.1
numba	0.50.1
numpy	1.19.5
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	8.9.2.26
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	2.19.3
nvidia-nvjitlink-cu12	12.4.127
nvidia-nvtx-cu12	12.1.105
packaging	24.0
pandas	1.4.4
path	16.14.0
patsy	0.5.6
pip	23.3.1
plotly	5.20.0
psutil	5.9.8
py3nvml	0.2.7
pyparsing	3.1.2
Pyro4	4.82
python-dateutil	2.9.0.post0
pytz	2024.1
PyYAML	6.0.1
scikit-learn	1.3.2
scipy	1.10.1
serpent	1.41
setuptools	68.2.2
six	1.16.0
SQLAlchemy	1.4.26
statsmodels	0.14.1
sympy	1.12
tenacity	8.2.3
texttable	1.7.0
threadpoolctl	3.4.0
torch	2.2.2
torchviz	0.0.2
tqdm	4.66.2
triton	2.2.0
typeguard	4.2.1
typing_extensions	4.11.0
wheel	0.41.2
xmltodict	0.13.0
zipp	3.18.1