🐈 Task: Refactor models with run.sh script

[kurysauce]

Summary

The updated test script reveals models that do not have an existing run.sh script. @GemmaTuron and @miquelduranfrigola agree that the solution is to refactor models to include run.sh. Running the update test script on a model will yield the error "Check halted. Either run.sh file does not exist, or model was not fetched via --from_github or --from_s3.", if a model does not already have a run.sh file.

Objective(s)

Refactor models to include a run.sh file. Models eos4e40 and eos3b5e do not have a run.sh file

Documentation

No response

[miquelduranfrigola]

Thank you @kurysauce !

For now, I am tagging here @HarmonySosa, @LauraGomezjurado and @DhanshreeA and we'll see who has capacity to jump into this issue.

Two comments:

1. eos3b5e should be easy to refactor and perhaps @HarmonySosa can help here, since she is dealing with it in the context of ersilia-pack?
2. eos4e40 should not be too difficult to refactor either, and perhaps we can gain inspiration from a related model eos3804?

[DhanshreeA]

Also adding @Malikbadmus and @dzumii here.

[dzumii]

Ok, checking this out

[kurysauce]

Currently working on eos4e40 with @miquelduranfrigola

[kurysauce]

Log on local bash run for eos4e40 shows that outputs are being captured.

(eos4e40) @kurysauce ➜ /workspaces/eos4e40/model/framework (main) $ bash run.sh . /workspaces/eos4e40/model/framework/input.csv /workspaces/eos4e40/model/framework/pred.csv
Reading file /workspaces/eos4e40/model/framework/input.csv
1 molecules read!
Running model!
content after reading the file: ['CC(C)(C)c1cc(O)c(cc1O)C(C)(C)C']
['python /workspaces/eos4e40/model/framework/code/save_features.py --data_path /workspaces/eos4e40/model/framework/data.csv --save_path /workspaces/eos4e40/model/framework/features.npz --features_generator rdkit_2d_normalized', 'python /workspaces/eos4e40/model/framework/code/predict.py --test_path /workspaces/eos4e40/model/framework/data.csv --checkpoint_dir /workspaces/eos4e40/model/checkpoints/final_model --preds_path /workspaces/eos4e40/model/framework/pred.csv --features_path /workspaces/eos4e40/model/framework/features.npz --no_features_scaling']
Calculations done
this is the result: [{'50uM_Inhibition': 0.0025936177141886673}]
each item is:{'50uM_Inhibition': 0.0025936177141886673}
header is none: so it is dict_keys(['50uM_Inhibition'])
appending header: dict_values([0.0025936177141886673])
Writing results to /workspaces/eos4e40/model/framework/pred.csv

pred.csv:
['50uM_Inhibition'],[0.0025936177141886673]

I also modified the code in processing the output and reading the smile input, not exactly sure on the expected data structure that the Ersilia command is looking for:

values = []
header = None
print(f"this is the result: {result}")
for item in result: 
    print(f"each item is:{item}")
    if header is None:
        print(f"header is none: so it is {item.keys()}")
        header = list(item.keys())
        print(f"appending header: {item.values()}")
        values.append(header) 
    values.append(list(item.values()))  

print("Writing results to", output_file)
with open(output_file, "w") as f:
    writer = csv.writer(f)
    writer.writerow(values)

[kurysauce]

Log on local bash run for eos4e40 shows that outputs are being captured.

(eos4e40) @kurysauce ➜ /workspaces/eos4e40/model/framework (main) $ bash run.sh . /workspaces/eos4e40/model/framework/input.csv /workspaces/eos4e40/model/framework/pred.csv
Reading file /workspaces/eos4e40/model/framework/input.csv
1 molecules read!
Running model!
content after reading the file: ['CC(C)(C)c1cc(O)c(cc1O)C(C)(C)C']
['python /workspaces/eos4e40/model/framework/code/save_features.py --data_path /workspaces/eos4e40/model/framework/data.csv --save_path /workspaces/eos4e40/model/framework/features.npz --features_generator rdkit_2d_normalized', 'python /workspaces/eos4e40/model/framework/code/predict.py --test_path /workspaces/eos4e40/model/framework/data.csv --checkpoint_dir /workspaces/eos4e40/model/checkpoints/final_model --preds_path /workspaces/eos4e40/model/framework/pred.csv --features_path /workspaces/eos4e40/model/framework/features.npz --no_features_scaling']
Calculations done
this is the result: [{'50uM_Inhibition': 0.0025936177141886673}]
each item is:{'50uM_Inhibition': 0.0025936177141886673}
header is none: so it is dict_keys(['50uM_Inhibition'])
appending header: dict_values([0.0025936177141886673])
Writing results to /workspaces/eos4e40/model/framework/pred.csv

pred.csv:
['50uM_Inhibition'],[0.0025936177141886673]

I also modified the code in processing the output and reading the smile input, not exactly sure on the expected data structure that the Ersilia command is looking for:

values = []
header = None
print(f"this is the result: {result}")
for item in result: 
    print(f"each item is:{item}")
    if header is None:
        print(f"header is none: so it is {item.keys()}")
        header = list(item.keys())
        print(f"appending header: {item.values()}")
        values.append(header) 
    values.append(list(item.values()))  

print("Writing results to", output_file)
with open(output_file, "w") as f:
    writer = csv.writer(f)
    writer.writerow(values)

resolved

[miquelduranfrigola]

Thanks @kurysauce this is awesome! Anytihng specific you need from me at this stage?

[kurysauce]

Thanks @kurysauce this is awesome! Anytihng specific you need from me at this stage?

Nope! PR has been merged. Just waiting for the implementation from @HarmonySosa on model eos3b5e to close the issue.

[kurysauce]

CLOSING COMMMENTS:
Existing incorporated models need to be tested for structural issues in repositories or hidden errors in main.py script. Open new issues as needed.

• eos4tcc is a work in progress 🐛 Bug: Inconsistent Output Error with model eos4tcc #1221
• eos3b5e work in progress PR
• eos4e40 has been refactored and tested (#11).
• eos8a5g has been tested and verified (string output)
• eos1pu1 has been tested and verified (float output)

[miquelduranfrigola]

Many thanks @kurysauce ! Very useful summary. We will take it from here. CC'in @DhanshreeA and @GemmaTuron

[DhanshreeA]

Update from @kurysauce 's work:

• eos3b5e has been merged.
• eos4tcc is still WIP. I'll take care of this, after which I will close this issue and 🐛 Bug: Inconsistent Output Error with model eos4tcc #1221 as well.

Many thanks again for the fantastic work Kurt!

[GemmaTuron]

I will close this issue as just the eos4tcc remains open with its own issue