google-deepmind/deepmind-research

This repository contains implementations and illustrative code to accompany DeepMind publications

deepmind-research's Issues

Questions on reproducing ScratchGAN results

Hi, thank you for keeping the code open-sourced! I am wondering if there is something I missed to reproduce the results of scratchGAN on EMNLP News dataset. The optimal FED is 0.0214 whereas that in the original paper was 0.015.

The results on single NVIDIA Titan RTX are summarized as follows:

checkpoint	mean train prob	mean valid prob	mean gen prob	FED
scratchgan-392002	0.0234371495898813	0.021976527175866067	0.016274933121167123	0.02389448
scratchgan-400000	0.022109536337666214	0.021964834304526448	0.015994238085113466	0.02222306
scratchgan-394002	0.023769843741320074	0.022377881105057895	0.016788717941381037	0.023813704
scratchgan-398002	0.025539066176861525	0.024043583893217146	0.018643759249243885	0.02179827
scratchgan-396002	0.024058140232227743	0.022415030980482697	0.016733897908125073	0.021409184

The hyper-parameter settings are default in the source code and the same as those reported in Appendix D.2. The num_steps was set to 200,000 by default.

Looking forward to your reply;)

Hi will add more information about memo project

i find the project only have detial about loading data

From distogram to structure?

Disclaimer: I'm not a structural biologist and I'm not working in protein structure prediction. I can't say I've understood all the details here.

I've read the the AlphaFold publication and was so intrigued by it that I wanted to predict the structure a couple of proteins that we typically look at (here in the lab) just out of curiosity.

From the README, I can see that it is possible to recreate the CASP13 results (distograms etc.).

My question: Is it possible to get from input protein sequence to predicted structure using AlphaFold?

Learning Mesh-Based Simulation with Graph Networks Code

Do you plan to release code for MeshGraphNets (Learning Mesh-Based Simulation with Graph Networks
https://arxiv.org/pdf/2010.03409.pdf) in the near future?

Thanks for awesome work.

Update of code for Graph Matching Networks to TF 2 planned?

Hi altogether, we are actively using the approach of Graph Matching Networks in research. Since version 2 of Tensorflow is now the main version and we want to make sure that our code base is future-proof, I was wondering whether an update of the GMN code to TF 2 and the respective version of Sonnet is planned? If not, I am also happy to hear hints on what has to be adjusted.

I hope one of the authors or any other affiliated person can give me a quick answer.

Thank you very much and best regards!

Datasets about Graph Matching Network

Hello, Diego, thanks for releasing the code of GMN! I find the dataset used in code is generated by networkx, which is simple to control the similarity of a pair or triplet of graphs. However, how to compute ground truth similarity of graphs in the other datasets used in paper, such as the control-flow graphs and COIL-DEL? Further more, I notice that the graph pair in GMN is given a similarity label(positive/negative) as ground truth, so how to transfer graph similarity value into label, is there a threshold to determine the graph pair is similar or not? I would appreciate it if you could answer my question. Best wishes~ :)

how to resolve this

File "/home/geethu/anaconda3/lib/python3.8/runpy.py", line 194, in _run_module_as_main
return _run_code(code, main_globals, None,
return _run_code(code, main_globals, None,
return _run_code(code, main_globals, None,
return _run_code(code, main_globals, None,
File "/home/geethu/anaconda3/lib/python3.8/runpy.py", line 87, in _run_code
return _run_code(code, main_globals, None,

wrong package in requirements.txt, missing aqs-data and improvement of tests

Hi everyone,

pkg-resources==0.0.0 can be found in the reuqirements.txt and causes an error when using pip install. According to this

https://stackoverflow.com/questions/39577984/what-is-pkg-resources-0-0-0-in-output-of-pip-freeze-command

this package does not exist and is a bug stemming from the operating system.
Also, I could not find the data of the athermal quasistatic simulations in the bucket. Could you please upload this?
When doing the tests and some fail or are skipped, there is no message indicating which test failed and what the implications are.

Regards
Stefan Hiemer

rl_unplugged/rwrl_d4pg.ipynb does not reproduce

The notebook is easy to get running, kudos for that. However the results do not match the repository.

When I run it the output of "Training Loop" is:

[Learner] Critic Loss = 4.062 | Policy Loss = 0.500 | Steps = 1 | Walltime = 0
[Learner] Critic Loss = 3.844 | Policy Loss = 0.269 | Steps = 46 | Walltime = 3.173
[Learner] Critic Loss = 3.770 | Policy Loss = 0.296 | Steps = 92 | Walltime = 4.182

and the "Evaluation":

[Evaluation] Episode Length = 1000 | Episode Return = 68.235 | Episodes = 1 | Steps = 1000 | Steps Per Second = 420.795
[Evaluation] Episode Length = 1000 | Episode Return = 73.514 | Episodes = 2 | Steps = 2000 | Steps Per Second = 448.120
[Evaluation] Episode Length = 1000 | Episode Return = 71.517 | Episodes = 3 | Steps = 3000 | Steps Per Second = 463.122
[Evaluation] Episode Length = 1000 | Episode Return = 74.285 | Episodes = 4 | Steps = 4000 | Steps Per Second = 464.442
[Evaluation] Episode Length = 1000 | Episode Return = 72.500 | Episodes = 5 | Steps = 5000 | Steps Per Second = 459.378

Is this expected?

ModuleNotFoundError

I don't know how to go about this issue. Can you please tell me what i'm doing wrong?

Errors about glassy_dynamics

Hi, I am interested in this paper, and thank you for your code.
I got errors when I run the code. Can you give me help about that? Thanks very much.
When I run the train_test.py, it shows that :

ERROR: test_apply_model (main.TensorflowTrainTest)
Tests if we can apply a model to a small test dataset.

Traceback (most recent call last):
File "/home/ymt1957/glassy_dynamics/train_test.py", line 130, in test_apply_model
time_index=0)
File "/home/ymt1957/glassy_dynamics/train.py", line 351, in apply_model
saver = tf.train.import_meta_graph(checkpoint_path + '.meta')
File "/home/ymt1957/anaconda3/envs/glass_dynamics/lib/python3.5/site-packages/tensorflow/python/training/saver.py", line 1435, in import_meta_graph
meta_graph_or_file, clear_devices, import_scope, **kwargs)[0]
File "/home/ymt1957/anaconda3/envs/glass_dynamics/lib/python3.5/site-packages/tensorflow/python/training/saver.py", line 1457, in _import_meta_graph_with_return_elements
**kwargs))
File "/home/ymt1957/anaconda3/envs/glass_dynamics/lib/python3.5/site-packages/tensorflow/python/framework/meta_graph.py", line 806, in import_scoped_meta_graph_with_return_elements
return_elements=return_elements)
File "/home/ymt1957/anaconda3/envs/glass_dynamics/lib/python3.5/site-packages/tensorflow/python/util/deprecation.py", line 507, in new_func
return func(*args, **kwargs)
File "/home/ymt1957/anaconda3/envs/glass_dynamics/lib/python3.5/site-packages/tensorflow/python/framework/importer.py", line 399, in import_graph_def
_RemoveDefaultAttrs(op_dict, producer_op_list, graph_def)
File "/home/ymt1957/anaconda3/envs/glass_dynamics/lib/python3.5/site-packages/tensorflow/python/framework/importer.py", line 159, in _RemoveDefaultAttrs
op_def = op_dict[node.op]
KeyError: 'SelectV2'

---

Ran 8 tests in 7.955s

FAILED (errors=1, skipped=2)

Process finished with exit code 1

alphafold_casp13 : the shape of profile_with_prior_without_gap & profile_with_prior

The shape of PSSM generated by PSI-BLAST with BLOUM62 is (NR, 20), there is no X and gap lines. How could you get the shape with (NR, 21) and (NR,22)?

curl for omniglot not working

When I ran
PYTHONPATH=pwd/..:$PYTHONPATH python3 train_main.py --dataset='omniglot'
It gave the following error:
Traceback (most recent call last): File "train_main.py", line 79, in <module> app.run(main) File "/data1/dhanajit/anaconda3/envs/curl/lib/python3.6/site-packages/absl/app.py", line 299, in run _run_main(main, args) File "/data1/dhanajit/anaconda3/envs/curl/lib/python3.6/site-packages/absl/app.py", line 250, in _run_main sys.exit(main(argv)) File "train_main.py", line 75, in main use_supervised_replay=False, File "/data1/dhanajit/Projects/2020/deepmind-research/curl/training.py", line 555, in run_training n_concurrent_classes, image_key, label_key) File "/data1/dhanajit/Projects/2020/deepmind-research/curl/training.py", line 129, in get_data_sources **dataset_kwargs) TypeError: FunctionWrapper object got multiple values for keyword argument 'split

Why am I there so many errors? I can't even run the run_eval without running into all these issues

curl for omniglot stuck after Iteration 1999

When running
PYTHONPATH=pwd/..:$PYTHONPATH python3 train_main.py --dataset='omniglot'
the code is stuck after iteration 2000:

I0301 12:30:59.622419 140350619187008 training.py:1115] Iteration 1999, Train batch purity: 1.000, Train elbo: 3810.632, Train kl y: 2.944, Train kl z: 67.379 n_active: 19
2020-03-01 12:31:08.276665: W ./tensorflow/core/framework/model.h:213] Encountered a stop event that was not preceded by a start event.

Is there some problem with the code or it takes too much time? It has been stuck for over 2 days now.

Also, is the purity same as the cluster accuracy which is mentioned in the paper?
Thank you.

BYOL convert weights to PyTorch

Hi,

Thanks for open sourcing this. It is very helpful. I am in the process of converting the BYOL R50x1 weights to PyTorch. I have been able to get the dimensions of the weights to match with the standard torchvision R50 model. When I evaluate the pytorch weights, I get ~70% on ImageNet val set. Any idea what I may be missing? I not sure, but 'SAME' padding in conv and max pool are primary suspects right now. Although it looks normal to me, is there any caveat in input image pre-processing?

error when running run_eval.sh: ModuleNotFoundError: No module named 'alphafold_casp13

Good Afternoon. I'm a biochem student doing some remote research while in quarantine. My prof suggested I use deepmind to study some proteins. I've been going to the process but I'm running into errors when running the run_eval.sh file command line in my ubuntu environment. The error log is below. I'm attaching the run_eval.sh file I'm using and a few screenshots. I hope you can help me. Thank you. Diego Torrejon. [
deep_mind error and run_eval file.zip

](url)

(alphafold) d@DESKTOP-OQPEK4U:/mnt/d/alphafold/deepmind-research-master$ bash /mnt/d/alphafold/deepmind-research-master/deepmind-research-master/alphafold_casp13/run_eval.sh
Requirement already satisfied: wheel in ./alphafold_venv/lib/python3.6/site-packages
Saving output to /home/d/contacts_T1019s2_2020_07_10_14_53_16/
Launching all models for replica 0
Launching all models for replica 1
Launching all models for replica 2
Launching all models for replica 3
Troubleshoot mark 1
All models running, waiting for them to complete
/mnt/d/alphafold/deepmind-research-master/alphafold_venv/bin/python3: Error while finding module specification for 'alphafold_casp13.contacts' (ModuleNotFoundError: No module named 'alphafold_casp13')
/mnt/d/alphafold/deepmind-research-master/alphafold_venv/bin/python3: Error while finding module specification for 'alphafold_casp13.contacts' (ModuleNotFoundError: No module named 'alphafold_casp13')
/mnt/d/alphafold/deepmind-research-master/alphafold_venv/bin/python3: Error while finding module specification for 'alphafold_casp13.contacts' (ModuleNotFoundError: No module named 'alphafold_casp13')
/mnt/d/alphafold/deepmind-research-master/alphafold_venv/bin/python3: Error while finding module specification for 'alphafold_casp13.contacts' (ModuleNotFoundError: No module named 'alphafold_casp13')
/mnt/d/alphafold/deepmind-research-master/alphafold_venv/bin/python3: Error while finding module specification for 'alphafold_casp13.contacts' (ModuleNotFoundError: No module named 'alphafold_casp13')
/mnt/d/alphafold/deepmind-research-master/alphafold_venv/bin/python3: Error while finding module specification for 'alphafold_casp13.contacts' (ModuleNotFoundError: No module named 'alphafold_casp13')
/mnt/d/alphafold/deepmind-research-master/alphafold_venv/bin/python3: Error while finding module specification for 'alphafold_casp13.contacts' (ModuleNotFoundError: No module named 'alphafold_casp13')
/mnt/d/alphafold/deepmind-research-master/alphafold_venv/bin/python3: Error while finding module specification for 'alphafold_casp13.contacts' (ModuleNotFoundError: No module named 'alphafold_casp13')
/mnt/d/alphafold/deepmind-research-master/alphafold_venv/bin/python3: Error while finding module specification for 'alphafold_casp13.contacts' (ModuleNotFoundError: No module named 'alphafold_casp13')
Troubleshoot mark 2
Ensembling all replica outputs
/mnt/d/alphafold/deepmind-research-master/alphafold_venv/bin/python3: Error while finding module specification for 'alphafold_casp13.ensemble_contact_maps' (ModuleNotFoundError: No module named 'alphafold_casp13')

How can I gain feature vectors or other middle outputs?

Hello, I am trying to do some work on the "Unsupervised Adversarial Training" cifar10 pre-train model. But I meet some problem that I do not know how to get the middle outputs besides the final logits output. Can you give me some advice on how to achieve my goal?

Thanks!

Cannot find implementation of Algorithm 1 and sensing matrices

I have read the Deep Compressed Sensing paper and the TensorFlow implementation of it in the csgan folder, but I was not able to find the implementation of Algorithm 1 from the paper. I want to reproduce the results of Table 2 and 3, but I cannot find a sensing matrix anywhere in the implementation, I can only find a neural network being used to measure the signal. Could you let me know where can I find the random sensing matrix and how can I reproduce the results reported in Table 2 and 3?

AlphaFold: Usage for single sequence?

As AlphaFold is intended to predict structure from a sequence, I want to input [some sequence of letters denoting an arbitrary amino acid sequence], and receive [a PDB file of the predicted structure(s)] as output. Can some part of the released code do this, using the released precomputed features? It is unclear from the readme how to use this code beyond replicating the CASP13 results.

For example, this modelling server does exactly that, but by method of comparing sequence similarity to existing homology models:
https://gomodo.grs.kfa-juelich.de/php/begin.php

I was told to post this here by the maintainer.

alphafold feature generation

Hello,
I'm interested in the details about feature generation described in the method section of the paper. Is there a plan to open source the script for generating the features from MSA data? or is it already exist, just that I didn't see it.

Thanks!

alphafold_casp13 on gpu?

I have replaced pip install tensorflow-gpu after the pip install -r requirement.txt step

(alphafold) [tru@sillage alphafold]$ pip list
Package                Version    
---------------------- -----------
absl-py                0.8.1      
astor                  0.8.1      
cloudpickle            1.3.0      
contextlib2            0.6.0.post1
decorator              4.4.1      
dm-sonnet              1.35       
gast                   0.3.3      
google-pasta           0.1.8      
grpcio                 1.27.2     
h5py                   2.10.0     
Keras-Applications     1.0.8      
Keras-Preprocessing    1.1.0      
Markdown               3.2.1      
numpy                  1.16.0     
pip                    20.0.2     
protobuf               3.11.3     
semantic-version       2.8.4      
setuptools             41.0.0     
six                    1.12.0     
tensorboard            1.14.0     
tensorflow             1.14.0     
tensorflow-estimator   1.14.0     
tensorflow-gpu         1.14.0     
tensorflow-probability 0.7.0      
termcolor              1.1.0      
Werkzeug               1.0.0      
wheel                  0.34.2     
wrapt                  1.12.0     

Now when I run the run_eval.sh script, I can see python3 on processes on the cpu/gpu but the gpu load is null.

Am I doing something wrong?

Thanks

I0226 15:18:26.553629 139620662380352 secstruct.py:87] Saving secstruct to /home/tru/virtualenv/Python3.6-centos7/alphafold/contacts_T1019s2_2020_02_26_15_01_11/distogram/3/secstruct/T1019s2-l64_s0.ss2
I0226 15:18:26.554534 139620662380352 secstruct.py:87] Saving secstruct to /home/tru/virtualenv/Python3.6-centos7/alphafold/contacts_T1019s2_2020_02_26_15_01_11/distogram/3/asa/T1019s2-l64_s0.ss2
I0226 15:18:26.555158 139620662380352 contacts.py:168] Evaluate 2 examples, 512 crops 256.0 crops/ex. Took 968.9s, 484.460 s/example 0.528 crops/s
Constructed T1019s2-l64_s0 len 64 from 256 chunks [64, 4 x 64, 4] in 438.9s
I0226 15:18:26.555906 139620662380352 contacts.py:191] SepWorking on 2 T1019s2 T1019s2 88
I0226 15:18:26.555965 139620662380352 contacts.py:194] Getting residue_index from features
I0226 15:18:37.838049 139620662380352 contacts.py:266] Constructed T1019s2-l64_s0 len 64 from 256 chunks [64, 4 x 64, 4] in 450.5s
I0226 15:18:37.838176 139620662380352 contacts.py:267] prob_accum[:, :, 1]: [[27.125   28.15625 29.1875  ... 24.875   23.84375 22.8125 ]
 [28.15625 29.375   30.40625 ... 25.90625 24.875   23.65625]
 [29.1875  30.40625 31.625   ... 26.9375  25.71875 24.5    ]
 ...
 [24.875   25.90625 26.9375  ... 23.      21.96875 20.9375 ]
 [23.84375 24.875   25.71875 ... 21.96875 21.125   20.09375]
 [22.8125  23.65625 24.5     ... 20.9375  20.09375 19.25   ]]
I0226 15:18:37.844177 139620662380352 secstruct.py:87] Saving secstruct to /home/tru/virtualenv/Python3.6-centos7/alphafold/contacts_T1019s2_2020_02_26_15_01_11/distogram/2/secstruct/T1019s2-l64_s0.ss2
I0226 15:18:37.845071 139620662380352 secstruct.py:87] Saving secstruct to /home/tru/virtualenv/Python3.6-centos7/alphafold/contacts_T1019s2_2020_02_26_15_01_11/distogram/2/asa/T1019s2-l64_s0.ss2
I0226 15:18:37.845623 139620662380352 contacts.py:168] Evaluate 2 examples, 512 crops 256.0 crops/ex. Took 977.7s, 488.836 s/example 0.524 crops/s
Constructed T1019s2-l64_s0 len 64 from 256 chunks [64, 4 x 64, 4] in 450.5s
I0226 15:18:37.848766 139620662380352 contacts.py:191] SepWorking on 2 T1019s2 T1019s2 88
I0226 15:18:37.848917 139620662380352 contacts.py:194] Getting residue_index from features

Every 1.0s: nvidia-smi                                                                                                                                                                     Wed Feb 26 15:23:56 2020

Wed Feb 26 15:23:56 2020
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 440.59	  Driver Version: 440.59       CUDA Version: 10.2     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  GeForce RTX 208...  Off  | 00000000:09:00.0  On |                  N/A |
| 26%   40C    P8    14W / 250W |    787MiB /  7981MiB |      6%      Default |
+-------------------------------+----------------------+----------------------+

+-----------------------------------------------------------------------------+
| Processes:                                                       GPU Memory |
|  GPU       PID   Type   Process name                             Usage      |
|=============================================================================|
|    0      4717      G   /usr/bin/X                                   114MiB |
|    0      5462      G   /usr/bin/gnome-shell                          93MiB |
|    0     27775      C   python3                                      115MiB |
|    0     27777      C   python3                                      115MiB |
|    0     27779      C   python3                                      115MiB |
|    0     27783      C   python3                                      115MiB |
|    0     27785      C   python3                                      115MiB |
+-----------------------------------------------------------------------------+

top:
  PID USER      PR  NI    VIRT    RES    SHR S  %CPU %MEM     TIME+ COMMAND                                                                                                                                        
27775 tru       20   0   13.5g   2.5g 175304 S 512.5  4.0  99:49.97 contacts.py                                                                                                                                    
27783 tru       20   0   13.5g   2.5g 175548 S 493.8  4.0 100:53.24 contacts.py                                                                                                                                    
27779 tru       20   0   13.5g   2.5g 175548 S 462.5  4.0 100:06.44 contacts.py                                                                                                                                    
27785 tru       20   0   13.5g   2.5g 175548 S 387.5  4.0  95:45.10 contacts.py                                                                                                                                    
27777 tru       20   0   13.5g   2.5g 175532 S 375.0  4.0  99:36.74 contacts.py                                                                                                                                    
 2157 root       0 -20       0      0      0 S   6.2  0.0   0:13.48 spl_dynamic_tas        

BYOL Performance with 40 epoch pretraining

https://github.com/deepmind/deepmind-research/tree/master/byol#running-a-short-pre-training-locally

In the case of byol trained on 40 epochs, what performance can you expect from linear evaluation?

Memory error

After running main.py with all dependencies installed, I get a MemoryError

I0526 16:25:56.645768 140080367834880 session_manager.py:502] Done running local_init_op.
Traceback (most recent call last):
File "main.py", line 200, in
app.run(main)
File "/home/leonarda/.local/lib/python2.7/site-packages/absl/app.py", line 300, in run
_run_main(main, args)
File "/home/leonarda/.local/lib/python2.7/site-packages/absl/app.py", line 251, in _run_main
sys.exit(main(argv))
File "main.py", line 180, in main
with tf.train.MonitoredSession(hooks=hooks) as sess:
File "/home/leonarda/.local/lib/python2.7/site-packages/tensorflow_core/python/training/monitored_session.py", line 1014, in init
stop_grace_period_secs=stop_grace_period_secs)
File "/home/leonarda/.local/lib/python2.7/site-packages/tensorflow_core/python/training/monitored_session.py", line 725, in init
self._sess = _RecoverableSession(self._coordinated_creator)
File "/home/leonarda/.local/lib/python2.7/site-packages/tensorflow_core/python/training/monitored_session.py", line 1207, in init
_WrappedSession.init(self, self._create_session())
File "/home/leonarda/.local/lib/python2.7/site-packages/tensorflow_core/python/training/monitored_session.py", line 1212, in _create_session
return self._sess_creator.create_session()
File "/home/leonarda/.local/lib/python2.7/site-packages/tensorflow_core/python/training/monitored_session.py", line 885, in create_session
hook.after_create_session(self.tf_sess, self.coord)
File "/home/leonarda/.local/lib/python2.7/site-packages/tensorflow_core/python/training/basic_session_run_hooks.py", line 572, in after_create_session
self._checkpoint_dir, "graph.pbtxt")
File "/home/leonarda/.local/lib/python2.7/site-packages/tensorflow_core/python/framework/graph_io.py", line 72, in write_graph
graph_def, float_format=''))
File "/home/leonarda/.local/lib/python2.7/site-packages/google/protobuf/text_format.py", line 197, in MessageToString
printer.PrintMessage(message)
File "/home/leonarda/.local/lib/python2.7/site-packages/google/protobuf/text_format.py", line 462, in PrintMessage
self.PrintField(field, element)
File "/home/leonarda/.local/lib/python2.7/site-packages/google/protobuf/text_format.py", line 557, in PrintField
self.PrintFieldValue(field, value)
File "/home/leonarda/.local/lib/python2.7/site-packages/google/protobuf/text_format.py", line 604, in PrintFieldValue
self._PrintMessageFieldValue(value)
File "/home/leonarda/.local/lib/python2.7/site-packages/google/protobuf/text_format.py", line 589, in _PrintMessageFieldValue
self.PrintMessage(value)
File "/home/leonarda/.local/lib/python2.7/site-packages/google/protobuf/text_format.py", line 454, in PrintMessage
self.PrintField(field, entry_submsg)
File "/home/leonarda/.local/lib/python2.7/site-packages/google/protobuf/text_format.py", line 557, in PrintField
self.PrintFieldValue(field, value)
File "/home/leonarda/.local/lib/python2.7/site-packages/google/protobuf/text_format.py", line 604, in PrintFieldValue
self._PrintMessageFieldValue(value)
File "/home/leonarda/.local/lib/python2.7/site-packages/google/protobuf/text_format.py", line 589, in _PrintMessageFieldValue
self.PrintMessage(value)
File "/home/leonarda/.local/lib/python2.7/site-packages/google/protobuf/text_format.py", line 464, in PrintMessage
self.PrintField(field, value)
File "/home/leonarda/.local/lib/python2.7/site-packages/google/protobuf/text_format.py", line 557, in PrintField
self.PrintFieldValue(field, value)
File "/home/leonarda/.local/lib/python2.7/site-packages/google/protobuf/text_format.py", line 604, in PrintFieldValue
self._PrintMessageFieldValue(value)
File "/home/leonarda/.local/lib/python2.7/site-packages/google/protobuf/text_format.py", line 589, in _PrintMessageFieldValue
self.PrintMessage(value)
File "/home/leonarda/.local/lib/python2.7/site-packages/google/protobuf/text_format.py", line 464, in PrintMessage
self.PrintField(field, value)
File "/home/leonarda/.local/lib/python2.7/site-packages/google/protobuf/text_format.py", line 557, in PrintField
self.PrintFieldValue(field, value)
File "/home/leonarda/.local/lib/python2.7/site-packages/google/protobuf/text_format.py", line 604, in PrintFieldValue
self._PrintMessageFieldValue(value)
File "/home/leonarda/.local/lib/python2.7/site-packages/google/protobuf/text_format.py", line 589, in _PrintMessageFieldValue
self.PrintMessage(value)
File "/home/leonarda/.local/lib/python2.7/site-packages/google/protobuf/text_format.py", line 464, in PrintMessage
self.PrintField(field, value)
File "/home/leonarda/.local/lib/python2.7/site-packages/google/protobuf/text_format.py", line 557, in PrintField
self.PrintFieldValue(field, value)
File "/home/leonarda/.local/lib/python2.7/site-packages/google/protobuf/text_format.py", line 622, in PrintFieldValue
out.write(text_encoding.CEscape(out_value, out_as_utf8))
File "/home/leonarda/.local/lib/python2.7/site-packages/google/protobuf/text_encoding.py", line 85, in CEscape
return ''.join(cescape_byte_to_str[ord(c)] for c in text)
MemoryError

I'm not sure what is the problem since I'm using everythig as it's supposed to be.

Error generated while executing

File "/home/geethu/anaconda3/lib/python3.8/runpy.py", line 194, in _run_module_as_main
Traceback (most recent call last):
File "/home/geethu/anaconda3/lib/python3.8/runpy.py", line 194, in _run_module_as_main
Traceback (most recent call last):
File "/home/geethu/anaconda3/lib/python3.8/runpy.py", line 194, in _run_module_as_main
Traceback (most recent call last):
File "/home/geethu/anaconda3/lib/python3.8/runpy.py", line 194, in _run_module_as_main
Traceback (most recent call last):
File "/home/geethu/anaconda3/lib/python3.8/runpy.py", line 194, in _run_module_as_main
Traceback (most recent call last):
File "/home/geethu/anaconda3/lib/python3.8/runpy.py", line 194, in _run_module_as_main
Traceback (most recent call last):
File "/home/geethu/anaconda3/lib/python3.8/runpy.py", line 194, in _run_module_as_main
Traceback (most recent call last):
File "/home/geethu/anaconda3/lib/python3.8/runpy.py", line 194, in _run_module_as_main
return _run_code(code, main_globals, None,
return _run_code(code, main_globals, None,
return _run_code(code, main_globals, None,
return _run_code(code, main_globals, None,
File "/home/geethu/anaconda3/lib/python3.8/runpy.py", line 87, in _run_code
return _run_code(code, main_globals, None,
File "/home/geethu/anaconda3/lib/python3.8/runpy.py", line 87, in _run_code
File "/home/geethu/anaconda3/lib/python3.8/runpy.py", line 87, in _run_code
exec(code, run_globals)
File "/home/geethu/anaconda3/lib/python3.8/runpy.py", line 87, in _run_code
File "/home/geethu/deepmind_research/alphafold_casp13/contacts.py", line 21, in
exec(code, run_globals)
File "/home/geethu/deepmind_research/alphafold_casp13/contacts.py", line 21, in
from absl import app
ModuleNotFoundError: No module named 'absl'
exec(code, run_globals)
from absl import app
File "/home/geethu/deepmind_research/alphafold_casp13/contacts.py", line 21, in
ModuleNotFoundError: No module named 'absl'
from absl import app
exec(code, run_globals)
File "/home/geethu/deepmind_research/alphafold_casp13/contacts.py", line 21, in
from absl import app
ModuleNotFoundError: No module named 'absl'
ModuleNotFoundError: No module named 'absl'
return _run_code(code, main_globals, None,
File "/home/geethu/anaconda3/lib/python3.8/runpy.py", line 87, in _run_code
exec(code, run_globals)
File "/home/geethu/deepmind_research/alphafold_casp13/contacts.py", line 21, in
from absl import app
ModuleNotFoundError: No module named 'absl'
Traceback (most recent call last):
File "/home/geethu/anaconda3/lib/python3.8/runpy.py", line 194, in _run_module_as_main
return _run_code(code, main_globals, None,
File "/home/geethu/anaconda3/lib/python3.8/runpy.py", line 87, in _run_code
exec(code, run_globals)
return _run_code(code, main_globals, None,
File "/home/geethu/anaconda3/lib/python3.8/runpy.py", line 87, in _run_code
File "/home/geethu/anaconda3/lib/python3.8/runpy.py", line 87, in _run_code
exec(code, run_globals)
File "/home/geethu/deepmind_research/alphafold_casp13/contacts.py", line 21, in
return _run_code(code, main_globals, None,
File "/home/geethu/anaconda3/lib/python3.8/runpy.py", line 87, in _run_code
from absl import app
exec(code, run_globals)
ModuleNotFoundError: No module named 'absl'
File "/home/geethu/deepmind_research/alphafold_casp13/contacts.py", line 21, in
from absl import app
ModuleNotFoundError: No module named 'absl'
File "/home/geethu/deepmind_research/alphafold_casp13/contacts.py", line 21, in
from absl import app
ModuleNotFoundError: No module named 'absl'
exec(code, run_globals)
File "/home/geethu/deepmind_research/alphafold_casp13/contacts.py", line 21, in
from absl import app
ModuleNotFoundError: No module named 'absl'
Ensembling all replica outputs
Traceback (most recent call last):
File "/home/geethu/anaconda3/lib/python3.8/runpy.py", line 194, in _run_module_as_main
return _run_code(code, main_globals, None,
File "/home/geethu/anaconda3/lib/python3.8/runpy.py", line 87, in _run_code
exec(code, run_globals)
File "/home/geethu/deepmind_research/alphafold_casp13/ensemble_contact_maps.py", line 24, in
from absl import app
ModuleNotFoundError: No module named 'absl'

Created a folder - deepmind inside alphafold_casp13(dataset)and run_eval.sh file - Is it ok ? running from inside deepmind

tensorflow has no attribute GraphKeys

Hi, not urgent, but I am attempting to reproduce the results by running the 'run.sh' script and hitting the error 'tensorflow' has no attribute 'GraphKeys'.

I think this was handled in the quick_eval_cifar.py tensorflow.compat.v1 import but not in the clever_hans scripts from which the error originates. Not sure how to best proceed. Traceback as follows. Any help would be appreciated.

Traceback (most recent call last):
File "/home/jpfrancis/anaconda3/lib/python3.7/runpy.py", line 193, in _run_module_as_main
"main", mod_spec)
File "/home/jpfrancis/anaconda3/lib/python3.7/runpy.py", line 85, in _run_code
exec(code, run_globals)
File "/home/jpfrancis/Development/deepmind-research/unsupervised_adversarial_training/quick_eval_cifar.py", line 30, in
from cleverhans import attacks
File "/home/jpfrancis/anaconda3/lib/python3.7/site-packages/cleverhans/attacks/init.py", line 9, in
from cleverhans.attacks_tf import SPSAAdam, margin_logit_loss, TensorAdam
File "/home/jpfrancis/anaconda3/lib/python3.7/site-packages/cleverhans/attacks_tf.py", line 17, in
from cleverhans import utils_tf
File "/home/jpfrancis/anaconda3/lib/python3.7/site-packages/cleverhans/utils_tf.py", line 341, in
loss_collection=tf.GraphKeys.REGULARIZATION_LOSSES):
AttributeError: module 'tensorflow' has no attribute 'GraphKeys'

TypeError: Can't instantiate abstract class GraphEncoder with abstract methods _build

Hello, I am interested in your paper, and thank you very much for your open source.
An error was encountered while running the code，can you give me help about that?
Thanks very much.

TypeError Traceback (most recent call last)
in ()
22
23 tensors, placeholders, model = build_model(
---> 24 config, node_feature_dim, edge_feature_dim)
25
26 accumulated_metrics = collections.defaultdict(list)

in build_model(config, node_feature_dim, edge_feature_dim)
59 ValueError: if the specified model or training settings are not supported.
60 """
---> 61 encoder = GraphEncoder(**config['encoder'])
62 aggregator = GraphAggregator(**config['aggregator'])
63 if config['model_type'] == 'embedding':

TypeError: Can't instantiate abstract class GraphEncoder with abstract methods _build

alphafold_casp13 tensorflow.python.layers.normalization.BatchNormalization AssertionError: Bad argument number for Name: 3, expecting 4

running the plain cpu version I get a lot of:

WARNING:tensorflow:Entity <bound method BatchNormalization.call of <tensorflow.python.layers.normalization.BatchNormalization object at 0x7efb6e190940>> 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 BatchNormalization.call of <tensorflow.python.layers.normalization.BatchNormalization object at 0x7efb6e190940>>: AssertionError: Bad argument number for Name: 3, expecting 4

problem in parsing hmm_profile feaature file

hello ，when I use the hhblit generate the hhm file and use the code here to parse the hhm file ,but
obvisouly it cannot work ,for the format are bellow:

the code is

it did not match，can you give some help？

The requirements.txt conflicts with itself

The requirements file conflicts with itself.

ERROR: tensorflow-io 0.8.0 has requirement tensorflow<1.16.0,>=1.15.0, but you'll have tensorflow 1.14.0 which is incompatible.

https://github.com/deepmind/deepmind-research/blob/685e16355e39bd4a906fe7ed4749b9e7e5e82b65/scratchgan/requirements.txt#L8

Will the Taichi code to produce the ground-truth simulations be released for "Learning to Simulate Complex Physics with Graph Networks"?

If it is possible, I think it might be interesting to see how changes in some parameters in the ground truth simulation might impact the learning process. Thanks

Generation of the profile_with_prior in alphafold

Hi!
Thanks for publishing your code and article. It's really useful and interesting.
I'm working on features generation pipeline and I'm stuck with profile_with_prior feature. Can you please explain the formula in the referred article or how did you calculate this feature? As I understand we need to calculate the distribution of the amino acids for every residue (it will be Position Probability Matrix in fact), based on MSA , which we get from HHBlits and then made some calculations according to proposed formula?

The Processed WikiText103

Will it be convenient to publish the processed WikiText103 data set

GMN application

Hi author,
The work is very well. I want to use this work for question answering.
I have one question, when run codes: I use 200-d vector for representiong node and edge features. After running, I find the attention weights can not align nodes well when compute two graphs match. Like this:

I do not know where is it wrong. Is it features vector error or others ?

Thank you

ywsun

Pre Trained Models for RL Unplugged

I was just wondering if there are any plans to release the pretrained models by which the dataset was generated. ( or a pointer if it has already been released.

Off by one error in Polygen?

If we look at this line, we see that the input sequence is padded on the right.

And in order to generate the predictive distribution, we do this , which makes sense - the last element in the sequence is not used for prediction. But then in an autoregressive setting, I would assume that you would have element 0, a_0 predict a_1 and so on.

In the training script, we have

vertex_model_loss = -tf.reduce_sum( vertex_model_pred_dist.log_prob(vertex_model_batch['vertices_flat']) * vertex_model_batch['vertices_flat_mask'])

This does not make sense to me. This would mean we are using the current element to predict itself. This has zero generative power, right?
@saran-t @charlienash

Reproducing the Deep Compressed Sensing results

I could not find the code for testing or validation in the Deep Compressed Sensing paper TensorFlow implementation in the csgan folder. I want to reproduce the results reported in the paper in Table 2 and 3, but in the code I can only see results reported during training, and cannot find a separate testing or validation dataset either. Are the results reported in the tables testing results, and if so, is that included in the provided code?

alphafold_casp13: wrong flag for mutual_information , 'MI' was not used in code, but labeled "√" in feature list

wrong flag for mutual_information , I didn't found 'mutual_information' in code(only existed in FEATURES, but not been parsed), while it was labeled "Needed" in feature list.

Reproducing the results in the GMN paper

Hi, thanks for releasing the GMN code, which is really well written ! I am currently trying to use GMN, and when I run the sample code, I notice that the auc on validation is around 70%. Could u please give some hints about how to reproduce the results in the paper (i.e., auc around 90% for the GED task)?

Would increasing training step works? If so, how large should it be? Thanks !

How to get the experimental results of mnist test data in the cs_gan and Deep Compressed Sensing?

I have two questions
1：How to obtain the experimental results of mnist test set in the paper? Which .py to run?
2：How to understand the positive and negative signs of the experimental results in the paper? I don't understand what the standard deviation means in the paper.
Thank you very much

AlphaFold folding pipeline

How can you obtain the final (folded) protein 3D structure from the distance map/distogram? In the README it says:

pasted/: Contains distograms obtained from the ensembled distograms by pasting. An RR contact map file is computed from this pasted distogram. This is the final distogram that was used in the subsequent AlphaFold folding pipeline in CASP13

But where is this AlphaFold folding pipeline? If this pipeline is not published, are there alternatives? I did not find a satisfactory solution yet.

incorrect m3a format in alphafold

When describing the deletion_probability feature for alphafold, you specify the fact that you use the m3a format from hhblits.
To quote hhblits' github on the m3a format:

residues emitted by Match states of the HMM are in upper case, residues emitted by Insert states are in lower case and deletions are written -.

In the m3a format, the sequences of the MSA are not necessarily of equal length, and deletions are denoted by "-", whereas lowercase letters denote insertions and cause the disparities in sequence length:

A B C        Original sequence
A - E        Sequence where residue 2 was deleted, residue 3 was substituded
A d B C      Sequence where a residue d was inserted between residue 2 and 3. 
             Note that now the residue B no longer aligns with that of the original sequence.

This means your description of the deletion_probability feature makes no sense: not just should we count "-" rather than lowercase letters if we are looking for deletions, but aligning the residues by column makes no sense in the a3m format, since the lengths dont match.

Assuming that the name deletion_probability is not a misnomer, one has to instead remove all lowercase letters form the a3m MSA and then count the number of "-" per column to obtain the probability of a deletion of a particular residue in the MSA.

Is my reasoning here correct, or am I missing something important?

RL Unplugged: Baseline implementations?

I was looking for the implementations of your baselines since the paper says that lack of baselines make algorithmic comparisons difficult but could not find anything neither in this repo nor in the ACME repo. Did I miss something or you did not publish the baselines?

Bug in dm_control_suite_d4pg.ipynb

The example code provided for running D4PG with tasks from DeepMind Control Suite has a bug.

The dataset preprocess function should concatenate the observations/next observations dict so that a complete observation is passed to the agent.

def flatten_observation(observation):
    keys = sorted(six.iterkeys(observation))
    observation_arrays = [tf.reshape(observation[key], [-1]) for key in keys]
    return tf.concat(observation_arrays, 0)

def preprocess_fn(sample):                                                      
    o_tm1, a_tm1, r_t, d_t, o_t = sample.data[:5]
    o_tm1 = flatten_observation(o_tm1)
    o_t = flatten_observation(o_t)
    return replay_sample.ReplaySample(
        info=sample.info, data=(o_tm1, a_tm1, r_t, d_t, o_t))

def main(argv):
    task = dm_control_suite.ControlSuite("cartpole_swingup")

    environment = task.environment
    environment_spec = specs.make_environment_spec(environment)

    dataset = dm_control_suite.dataset(
        "./tmp",
        data_path=task.data_path,
        shapes=task.shapes,
        uint8_features=task.uint8_features,
        num_threads=1,
        batch_size=256,
        num_shards=100)

    dataset = dataset.map(preprocess_fn).batch(256)
    

Frobenius norm in AlphaFold pipeline

Dear AlphaFold team,

I have a question about the pseudo_frob feature in AlphaFold. This feature is supposed to be the Frobenius norm of the coupling matrix of each pair of sites in the Potts model. Since the coupling matrix of a site with itself is by definition, the zero matrix (as you can also verify checking the pseudolikelihood feature), I would expect the diagonal of the pseudo_frob features to be composed of zeros exclusively.

However, this is not what one finds in the published features. For example, for T0953s1:

Would you be able to clarify this?

Deep Compressed Sensing TensorFlow implementation missing CelebA dataset configurations

I have read the Deep Compressed Sensing paper and the TensorFlow implementation of it in the csgan folder. My goal is to reproduce the results from Table 2 and 3, and I have already added the sensing matrices to the code after posting a question here, and I was able to reproduce the results for the MNIST dataset (Table 2). However, the provided code does not include the CelebA dataset, so I cannot reproduce the results from Table 3 yet. I want to add the CelebA dataset, but I am not sure what the exact setup was to create the results in Table 3. What parameters were used for the dataset (most importantly, the image dimensions), and which generator and discriminator do I need to use to reproduce the results in Table 3?

Data links are not available.

Trying to download from the provided links on Readme.md is not working. Link is down or expired. Please, provide a valid link for the data.

gpu support for curl

Does CURL implementation have support GPU?
I am unable to run it entirely on GPU, and it is mostly running on the CPU.

Evaluating AlphaFold using Windows 10, WSL and Ubuntu

Hi, I'm a beginner trying to get AlphaFold to work in Windows 10.
I have WSL enabled and Ubuntu installed.

I tried to run the script with the requirements, and ran into problems, so I installed each python package individually in my virtual environment, and then commented out the packages in the requirements file, so that it would get past this stage.

I then ran the bash script in Windows Command Prompt using this command, and received the following error traceback in my command prompt.

`bash alphafold_casp13/run_eval.sh`


`(alphafold_env) C:\Users\....\deepmind-research-master>bash alphafold_casp13/run_eval.sh
Collecting wheel Using cached wheel-0.35.1-py2.py3-none-any.whl (33 kB)
Installing collected packages: wheel
Successfully installed wheel-0.35.1
Saving output to /home/user/contacts_T1019s2_2020_09_06_12_10_32/
Launching all models for replica 0
Launching all models for replica 1
Launching all models for replica 2
Launching all models for replica 3
All models running, waiting for them to complete
Traceback (most recent call last):
File "/usr/lib/python3.8/runpy.py", line 193, in _run_module_as_main
  return _run_code(code, main_globals, None,
File "/usr/lib/python3.8/runpy.py", line 86, in _run_code
  exec(code, run_globals)
Traceback (most recent call last):
Traceback (most recent call last):
File "/mnt/c/Users/..../deepmind-research-master/alphafold_casp13/contacts.py", line 21, in <module>
File "/usr/lib/python3.8/runpy.py", line 193, in _run_module_as_main
File "/usr/lib/python3.8/runpy.py", line 193, in _run_module_as_main
Traceback (most recent call last):
  return _run_code(code, main_globals, None,
  return _run_code(code, main_globals, None,
File "/usr/lib/python3.8/runpy.py", line 193, in _run_module_as_main
File "/usr/lib/python3.8/runpy.py", line 86, in _run_code
File "/usr/lib/python3.8/runpy.py", line 86, in _run_code
  return _run_code(code, main_globals, None,
exec(code, run_globals)
exec(code, run_globals)
File "/usr/lib/python3.8/runpy.py", line 86, in _run_code
File "/mnt/c/Users/..../deepmind-research-master/alphafold_casp13/contacts.py", line 21, in <module>
File "/mnt/c/Users/....g/deepmind-research-master/alphafold_casp13/contacts.py", line 21, in <module>
  exec(code, run_globals)
File "/mnt/c/Users/..../deepmind-research-master/alphafold_casp13/contacts.py", line 21, in <module>
Traceback (most recent call last):
File "/usr/lib/python3.8/runpy.py", line 193, in _run_module_as_main
  return _run_code(code, main_globals, None,
from absl import app
File "/usr/lib/python3.8/runpy.py", line 86, in _run_code
from absl import app
from absl import app
from absl import app
ModuleNotFoundError: No module named 'absl'
  exec(code, run_globals) 
ModuleNotFoundError: No module named 'absl'
ModuleNotFoundError: No module named 'absl'
Traceback (most recent call last):
ModuleNotFoundError: No module named 'absl'
  File "/mnt/c/Users/..../deepmind-research-master/alphafold_casp13/contacts.py", line 21, in <module>
Traceback (most recent call last):
Traceback (most recent call last):
  File "/usr/lib/python3.8/runpy.py", line 193, in _run_module_as_main
from absl import app
File "/usr/lib/python3.8/runpy.py", line 193, in _run_module_as_main
File "/usr/lib/python3.8/runpy.py", line 193, in _run_module_as_main
  return _run_code(code, main_globals, None,
ModuleNotFoundError: No module named 'absl'
  return _run_code(code, main_globals, None,
  return _run_code(code, main_globals, None,
File "/usr/lib/python3.8/runpy.py", line 86, in _run_code
File "/usr/lib/python3.8/runpy.py", line 86, in _run_code
File "/usr/lib/python3.8/runpy.py", line 86, in _run_code
  exec(code, run_globals)
Traceback (most recent call last):
  exec(code, run_globals)
  exec(code, run_globals)
File "/mnt/c/Users/..../deepmind-research-master/alphafold_casp13/contacts.py", line 21, in <module>
File "/usr/lib/python3.8/runpy.py", line 193, in _run_module_as_main
File "/mnt/c/Users/..../deepmind-research-master/alphafold_casp13/contacts.py", line 21, in <module>
File "/mnt/c/Users/..../deepmind-research-master/alphafold_casp13/contacts.py", line 21, in <module>
from absl import app
  return _run_code(code, main_globals, None,
from absl import app
ModuleNotFoundError: No module named 'absl'
  from absl import app
File "/usr/lib/python3.8/runpy.py", line 86, in _run_code
ModuleNotFoundError: No module named 'absl'
ModuleNotFoundError: No module named 'absl'
  exec(code, run_globals)
File "/mnt/c/Users/..../deepmind-research-master/alphafold_casp13/contacts.py", line 21, in <module>
from absl import app
ModuleNotFoundError: No module named 'absl'
Ensembling all replica outputs
Traceback (most recent call last):
File "/usr/lib/python3.8/runpy.py", line 193, in _run_module_as_main
return _run_code(code, main_globals, None,
File "/usr/lib/python3.8/runpy.py", line 86, in _run_code
exec(code, run_globals)
File "/mnt/c/Users/..../deepmind-research-master/alphafold_casp13/ensemble_contact_maps.py", line 24, in <module>
from absl import app
ModuleNotFoundError: No module named 'absl'

(alphafold_env) C:\....\deepmind-research-master>`

I have Python 3.6 installed on my machine. But I used to have python 3.8 installed. I uninstalled python 3.8 from my machine. So not sure why Python 3.8 is in the traceback.

(alphafold_env) C:\....\deepmind-research-master>python --version Python 3.6.4

Is it possible that when I run this run_eval bash script in windows command prompt, it finds the version of Python 3.8 that was installed in Windows, still in Ubuntu ?

I was wondering if anyone has a walkthrough for beginners to AlphaFold that use Windows 10 ?

Or if there are any other suggestions regarding how I could get AlphaFold running, that would be much appreciated.

In rl_unplugged, is your dependencies right?

The TF version in my Linux env is 2.2, i installed the dependencies based on your requirement.txt, when i run your atari demo, it note me that 'no moudle named tf.contrib', so i down the TF version to 1.15, and Then it note me that 'the version of Reverb require TF version >=2.3.0, and then i upgrade the TF to 2.3.0. Now it notes me 'no moudle named tf.contrib'