elijahcole / sinr Goto Github PK

Anaconda automatically installs libdeflate-1.17. Unfortunately this fails with a LIBDEFLATE_INSUFFICIENT_SPACE error when running format_env_feats.py. Downgrading to libdeflate-1.8 fixes this.

The instructions in data/README.md result in the data being downloaded and upzipped to data/data/. Need to update the instructions such that the directory structured specified in the readme to obtained.

Hi! I'm trying to build a similar network, but I would like to experiment with potential encodings besides the regular sin-cos version. It is of course possible to generate the raw data from the provided sin-cos version, but I'm a bit worried about accuracy issues. Can you please kindly open-source the raw data without the sin-cos encoding? Thank you!

Mirrors in the data section for environmental data are currently down:

curl -L https://biogeo.ucdavis.edu/data/worldclim/v2.1/base/wc2.1_5m_bio.zip --output wc2.1_5m_bio.zip curl -L https://biogeo.ucdavis.edu/data/worldclim/v2.1/base/wc2.1_5m_elev.zip --output wc2.1_5m_elev.zip

Have been down about a week. Is there anywhere else to get them from?

Hi! Thanks for creating and open-sourcing this awesome project. I was wondering something about testing details: how exactly are the points used for calculating MAP values sampled? If this was calculated using the hexagon grid, does it make the testing results biased towards negative results (since there are more negative regions than positive ones)? Thank you in advance for your answer.

Hi! Thanks again for the work, it really helped me a lot. However, I did notice some small inconveniences with the codebase, so I thought the codebase could be improved via fixing those.

1. np.int was used for 4 times in datasets.py. It was deprecated long ago and should be replaced with np.int32.
2. If experiments/demo exists the program would fail to run due to FileExists. It should be autodeleted when the code is ran, or saved under a name that includes the local time.
3. Training an epoch takes around 30 minutes on my server but there's no display of progress. I recommend adding a tqdm progress bar.

I thought it would be helpful to introduce those into the codebase. If the authors agree I could open a pull request with the fixes implemented ;-)

Hi it's me again :-)

When I was using the web app it complained for return op_html, fig, gr.Number.update(value=eval_params['taxa_id']) that AttributeError: type object 'Number' has no attribute 'update'.

I changed the line to return op_html, fig, eval_params['taxa_id'] and it seems to work fine.

Hi,

Thanks for your great work!

I am wondering will the dataset be released alongside the code? Or if there is some other way for us to obtain the dataset from https://www.inaturalist.org/?

Line 76 of the code produces a "shape mismatch" message. This appears to caused by preds.shape being too large (499385,256) to be pasted into op_im (shape: (499385,)). Which slice of preds is the correct one to use?

See line 69:
dsf = decomposition.FastICA(n_components=num_ds_dims, random_state=seed, whiten=True, max_iter=1000)

Produces this error:
sklearn.utils._param_validation.InvalidParameterError: The 'whiten' parameter of FastICA must be a str among {'arbitrary-variance', 'unit-variance'} or a bool among {False}. Got True instead.

Replace with:
dsf = decomposition.FastICA(n_components=num_ds_dims, random_state=seed, whiten='unit-variance', max_iter=1000)