Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Activity
    Abdul Karim
    @Abdulk084
    Here is another question with sample code and sample data. A very useful I guess , I would appreciate the answer. https://stackoverflow.com/questions/58295921/deepchem-disk-data-to-numpy-array
    Or some clues to inspect diskdata object.
    Mostafa Nategh
    @1particle

    Greetings, I'm running a graphconv classifier on a dataset gathered from chembl, and it runs ok. But when I try to run the same code on the HIV or Tox21 datasets (by manually loading the csv files, as that is my use case), featurization completes, but the fit function fails with:

    Traceback (most recent call last):
    File "graphconv.py", line 75, in <module>
    model.fit(train, nb_epoch=1, checkpoint_interval=0)
    File "/home/m/miniconda3/lib/python3.6/site-packages/deepchem/models/tensorgraph/tensor_graph.py", line 153, in fit
    max_checkpoints_to_keep, checkpoint_interval, restore, submodel)
    File "/home/m/miniconda3/lib/python3.6/site-packages/deepchem/models/tensorgraph/tensor_graph.py", line 231, in fit_generator
    for feed_dict in self._create_feed_dicts(feed_dict_generator, True):
    File "/home/m/miniconda3/lib/python3.6/site-packages/deepchem/models/tensorgraph/tensor_graph.py", line 1071, in _create_feed_dicts
    for d in generator:
    File "/home/m/miniconda3/lib/python3.6/site-packages/deepchem/models/tensorgraph/models/graph_models.py", line 704, in default_generator
    d[self.labels[0]] = to_one_hot(y_b.flatten(), self.n_classes).reshape(
    File "/home/m/miniconda3/lib/python3.6/site-packages/deepchem/metrics/init.py", line 30, in to_one_hot
    y_hot[np.arange(n_samples), y.astype(np.int64)] = 1
    IndexError: index 4 is out of bounds for axis 1 with size 2

    When I use a different column of csv, the index number in the last line of the error message changes. This seems so strange to me.
    I can submit the whole code file if needed.
    Your help is much appreciated, as always! :)

    Bharath Ramsundar
    @rbharath
    Could you try raising an issue on github with a minimal failing example?
    Usman Zafar
    @OsmanXafar_twitter
    Greetings!. I am getting the following error on google colab(courtesy @jhjensen2 ).
    AttributeError: module 'tensorflow.contrib.eager' has no attribute 'in_eager_mode'
    I was able to run my code successfully a month ago but now it is not working.
    Bharath Ramsundar
    @rbharath
    Could you check your TensorFlow version?
    Mostafa Nategh
    @1particle
    Will do. Thanks Bharath!
    Abdul Karim
    @Abdulk084
    What is the difference between Graph convolution and Weave model of deepchem? From the paper, it seems to me that Weave is a type of GC.
    John Thiels
    @jthiels_gitlab

    Hey, I've loaded deepchem in conda env on Ubuntu 18.04, get this error when running the nosetests test in the github instructions: nosetests -a '!slow' -v deepchem --nologcapture
    /usr/local/share/anaconda3/lib/python3.7/site-packages/sklearn/externals/joblib/init.py:15: DeprecationWarning: sklearn.externals.joblib is deprecated in 0.21 and will be removed in 0.23. Please import this functionality directly from joblib, which can be installed with: pip install joblib. If this warning is raised when loading pickled models, you may need to re-serialize those models with scikit-learn 0.21+.
    warnings.warn(msg, category=DeprecationWarning)
    Failure: KeyError ("Registering two gradient with name 'BlockLSTM'! (Previous registration was in register /usr/local/share/anaconda3/lib/python3.7/site-packages/tensorflow_core/python/framework/registry.py:66)") ... ERROR

    ======================================================================
    ERROR: Failure: KeyError ("Registering two gradient with name 'BlockLSTM'! (Previous registration was in register /usr/local/share/anaconda3/lib/python3.7/site-packages/tensorflow_core/python/framework/registry.py:66)")

    Bharath Ramsundar
    @rbharath
    @Abdulk084 Yes, weave is a type of graph conv, but the internals are quite different. I recommend checking out Google's paper to learn more
    @jthiels_gitlab This one is new to me. Can you check which version of deepchem/tensorflow you're on?
    John Thiels
    @jthiels
    @rbharath The deepchem was the master branch built from source; tensorflow updated to 2.0, python 3.5 (as per the deepchem install script) I kept running into other TF errors running deepchem with earlier versions of TF as well, specifically mentioning AutoGraph
    John Thiels
    @jthiels
    I am working in a singularity container sandbox with Ubuntu 18.04, have tried with Anaconda and Miniconda, in conda, from source, etc, and there seem to be constant issues with the nosetests...listed in the github repo--if I try the latest deepchem release and tensorflow 2.0 and python 3.7, will that give me a better chance at it working?
    abanerjee10
    @abanerjee10_gitlab
    I have built the graph conv model from https://github.com/deepchem/deepchem/blob/master/examples/notebooks/graph_convolutional_networks_for_tox21.ipynb, and I am trying to call compute_saliency and predict on it. However, I am not able to figure out what input to pass into the methods?
    abanerjee10
    @abanerjee10_gitlab
    More specifically, I thought that the dataset that I got in the tutorial from calling load_tox21 with the GraphConv featurizer would be correct input, but passing them in results in the error "float() argument must be a string or a number, not 'ConvMol'." I am using deepchem 2.3
    Bharath Ramsundar
    @rbharath
    @abanerjee10_gitlab Could you raise an issue on github with a minimal failing example?
    abanerjee10
    @abanerjee10_gitlab
    @rbharath done. I have raised an issue at deepchem/deepchem#1706
    Christian Stemmle
    @chstem
    Hi, what is the recommended way to install deepchem? I tried running conda install -c deepchem deepchem=2.1.0 python=3.6 as the website suggests, but I get a long list of conflicts. In a new conda environment using conda 4.7.12.
    Bharath Ramsundar
    @rbharath
    Ack looks like the website build failed so we have an old command still
    Try conda install -c deepchem -c rdkit -c conda-forge -c omnia deepchem=2.3.0
    I'll try to take a look at the website to see what's happening there
    Christian Stemmle
    @chstem
    @rbharath thank you, this works.
    are there any plans to support a more recent version of tensorflow?
    Bharath Ramsundar
    @rbharath
    Yes, the latest version 2.3 is up to TF 1.14
    We're working on the upgrade to TF 2 right now
    DeepChem's next release should be on TF 2, but this will take a bit of work (moving everything over to eager mode)
    yuanjames
    @yuanjames
    Hi, Could I ask where I can find the details of the database?
    abanerjee10
    @abanerjee10_gitlab
    Is it possible to change the weights of classes when training deepchem models? Is there a tutorial or method for doing this?
    genesistutorials
    @genesistutorials

    Studying Chemistry Opens Doors to a Range of Sectors and Opportunities
    Chemistry careers in research
    Chemistry careers in chemical engineering
    Chemistry careers in healthcare
    Chemistry careers in pharmaceuticals
    Chemistry careers in the public sector

    Join for JAM in Chemistry, CSIR-UGC-NET/JRF in Chemical Science & GATE in Chemistry. #CSIR_DECEMBER_2019 #GATE_2020 & #JAM_2020_EXAM in Chemistry.
    Contact for Regular & Weekend Both. Book Your seat.
    http://www.genesis-tutorials.com/
    Contact: 08171000425, 07252022576
    Email: info@genesis-tutorials.com

    CSIR_NETCoaching #GATE_Coaching #JAM_Coaching #JRF_Coaching #Smart_Classes #Knowledge #career_in_chemistry #Chemical_Science #Genesis_Tutorials #Build_Your_Career #Modern_Educator #Change_the_World #Keep_Learning

    Christian Stemmle
    @chstem
    about the tox21 tutorial: is there somewhere an explanation what the GraphGather layer actually does?
    vincenzodentamaro
    @vincenzodentamaro
    Hi all, I am a PhD student in pattern recognition in healthcare from University of Bari "Aldo Moro" Italy. I bought the book Deep Learning for the Life Sciences, but I am missing some major background knowledge in coputational chemistry. My specialization is Machine Learning, which is what I know, but not chemistry at this level.
    Can you please provide me some book or link where I can quickly catch up with terminology and understand why using these computational models?
    Thanks
    Jan H. Jensen
    @jhjensen2
    vincenzodentamaro
    @vincenzodentamaro
    thanks
    I am new to this and would like to learn
    Jan H. Jensen
    @jhjensen2
    The last two videos on this page might also be useful https://sites.google.com/view/ml-basics/convolution
    Bharath Ramsundar
    @rbharath
    @chstem The graph gather basically gathers the local atomic feature vectors from each atom and constructs a molecule-wide feature vector for the full molecule
    Iirc, it might just be summing the feature vectors, but I can't say off the top of my head
    @abanerjee10_gitlab I think the way you could do this would be to change the weights on the dataset after each epoch. There's not a tutorial on doing this, but given a Dataset object dataset, you need to alter dataset.w to have the new weights desired
    Tim Komm
    @TimKomm_twitter

    Hi I dont know if that is the correct spot to post this. But the tutorials on the deepchem (not github) website are not update to date for the GCN Tox21 code.

    Greetings Janosch

    しゃお
    @Jinniuzuo_dier_twitter

    Hi, I don't know it is correct to ask this here, but I want to know this.
    I want to execute this import, but I cannot import because of ModuleNotFoundError.

    import numpy as np
    import tensorflow as tf
    import deepchem as dc
    from deepchem.models.tensorgraph.models.graph_models import GraphConvModel

    error message is below

    ModuleNotFoundError                       Traceback (most recent call last)
    <ipython-input-27-a971180d73d7> in <module>
          2 import tensorflow as tf
          3 import deepchem as dc
    ----> 4 from deepchem.models.tensorgraph.models.graph_models import GraphConvModel
    
    ModuleNotFoundError: No module named 'deepchem.models.tensorgraph.models.graph_models'

    Could you please give me help?
    Thank you.

    Christian Stemmle
    @chstem
    @Jinniuzuo_dier_twitter apparently the website is not up-to-date with the library. Use from deepchem.models.graph_models import GraphConvModel instead
    しゃお
    @Jinniuzuo_dier_twitter
    @chstem san, thank you for confirmation and reply! I will try to import
    from deepchem.models.graph_models import GraphConvModel
    Jan H. Jensen
    @jhjensen2
    @OsmanXafar_twitter The solution is to use deepchem-gpu=2.3.0. I have updated the notebook
    ad05bzag
    @ad05bzag

    Hi, I installed Deepchem using
    !conda install -y -c deepchem -c rdkit -c conda-forge -c omnia deepchem-gpu=2.3.0
    Now i m trying to do seq2seq tutorial. I am getting stuck at creating the deepchem.models.Seq2Seq. It gets created with the following error:
    WARNING:tensorflow:Entity <bound method Stack.call of <deepchem.models.layers.Stack object at 0x7fb693186750>> could not be transformed and will be executed as-is. Please report this to the AutgoGraph team. When filing the bug, set the verbosity to 10 (on Linux, `export AUTOGRAPH_VERBOSITY=10`) and attach the full output. Cause: converting <bound method Stack.call of <deepchem.models.layers.Stack object at 0x7fb693186750>>: AssertionError: Bad argument number for Name: 3, expecting 4

    created model does not have any params, so
    model.get_params() = {'annealing_final_step': None, 'annealing_start_step': None, 'decoder_layers': None, 'dropout': None, 'embedding_dimension': None, 'encoder_layers': None, 'input_tokens': None, 'max_output_length': None, 'output_tokens': None, 'reverse_input': None, 'variational': None}

    tf version
    tf.__version__ = 1.14.0
    deepchem version
    dc.__version__ = 2.3.0