[Internship Project]: Simran Tandon #716
Comments
|
In this example, following the tasks of Wednesday 21st June:
I have created all the templates for the Interns, spend a couple of hours revising the GitHub Project and updating the tasks. I have been working on identifying the bug in the GitActions, solved in this issue. Pro tip: adding the links to the issues and discussions you mention will be very helpful! |
|
Wrote this in the wrong place! |
|
6/23
I completed testing on eos92sw on all three platforms. I also worked on fixing an issue with docker for a while (managed to get the model to run in docker itself, still having issues with the wsl-docker connection however). I dedicated a portion of time to examining the issues in eos2hbd further, looking at the log on actions and in CLI after running --repo_path on a local version and seeing where it starts to fail and found a possible specific line causing the issue. I then tried to find where to edit this line but could not see where it was in the models workflow or code files. I tried testing eos2v11 on CLI, Colab, and docker but on CLI and Colab ran into fetching issues. |
|
Hi @simrantan ! Thansk for the summary, I've added a few pointers to help with the issoes on eos2hbd |
|
Reflection 6/26
I tested eos22io on Docker, CLI, and Colab, finding no issues when I ran the model. I also finally solved the eos2hbd error and was able to fetch the model and test it. I refactored according to instructions and plan to test the changed version locally tomorrow and submit a PR if it does. |
|
Reflection 6/27
I was able to test eos24jm with no issues on all platforms. I also finished refactoring eos2hbd. The first refactor I did yesterday wasn't working when I tested locally, but after some bug hunting I found that I had created a syntax error when editing service.py that was preventing the model from being fetched successfully. I fixed the error, re-cloned and tested the model, and it worked on the inputs I tried! . |
|
I've added a couple of models to work on and some model testing! |
Tested eos46ev on CLI, Docker, Colab - ran into an issue with CLI when using eml_canonical but it ran fine w a single output. I tried to open an issue with one of the drafts suggestions I saw in the tasks by subprojects chart but I think I messed up by setting the repo to ersilia instead of the model's repo - I am working on fixing it, so sorry about that! Also, for future reference - if issues are drafts should I wait for them to be opened by you or Miquel or should I open it myself(properly, not like what I did this time haha) |
|
Hi @simrantan Apologies, I should have converted the draft to an issue, it is fixed now. |
I refactored eos3zur(not much needed, checked through everything against the guidelines and made what changes were needed). Tested it locally with varied inputs and submitted a PR! I also tested eos4zfy on Docker, ClI, and Colab. . Then, I tested eos6aun on Docker, CLI, and Colab successfully as well. Have started looking into the error in eos1amn but haven't found any way to fix it yet. |
|
@simrantan apologies for the eos1amn |
|
Awesome! Happy to help.
I ended up discovering an issue with my laptop today when working on eos4se9 (a storage overflow that had overwritten other data and caused a lot of applications to freak out and shut down) and ended up spending most of the day doing damage control on this issue (having to delete and reinstall many affected applications). I lost all the progress I made on eos4se9 as one of the side effects, but I have been working on restoring the changes I made. I will make sure to finish refactoring and submit a pr for both of these models next working day, I apologize for the unexpected delay! |
I worked with Febie to test eos1af5 on Windows and ended up finding a new possible source of issues in the model (some package or version issues?) through some attempts at fetching and running the model locally. I also refactored eos4se9, and tested it locally, finding a bug that I am worked on resolving. . It now works! I submitted a PR for this model. I refactored eos2rd8, and will test locally tomorrow. I also tested eos3cf4 on CLI and Colab, finding the same error as Hellen (module error) on both platforms. I also tested eos6pbf on CLI and Colab, and will test on Docker tomorrow. |
|
Hi @simrantan |
I have been working on testing the models I have refactored locally. I spent some time testing eos4se9, trying to implement print statements in the code to see if it is processing the inputs or if it is stalling for some reason. I am working on implementing the print statements, though I am having a little bit of trouble with getting the code to print to terminal.. I also tested eos2rd8 locally, found that it worked, and submitted a PR. I also tested eos3zur locally and submitted a PR for that as well. Docker is taking a long time to pull the image but hopefully I will be able to run it eos6pbf soon. |
|
Hi @simrantan |
I discovered that my docker desktop app had run into a WSL issue when installing the update, so I followed instructions to fix it and was able to test eos6pbf on docker successfully. . I also refactored and tested eos8h6g locally, and submitted aPR for it. Finally, I spent some time working to make sure eos4se9 was working on eml_canonical, finally confirmed that the model works, and submitted a PR. |
|
Perfect, I have assigned model testing for |
|
Thank you for the tip!
Today, I spent a good chunk of time looking at eos7pw8 back in time and the current version with Samuel's additions to see how he implemented the caching in the model. I also looked at some online resources to better understand caching in python, and *began identifying files that could be cached in eos2rd8. Tomorrow, I will implement the caching and test the model to ensure it still works. Is there a way to know the size of a model before and after caching a file? Or is it manual subtraction of the file size from the overall size? |
|
Thanks for the explanations @simrantan ! |
Committed my refactored version and submitted PR for eos5axz. Today, I worked on adding caching to the dockerhub workflow file for eos2rd8. I knew what parts I wanted to cache, and once I figured out more about the process of caching through the github documentation I was able to start writing the code, then working on the errors that arose with syntax and key mapping. I had a few questions about proceeding. Hopefully it works! Otherwise I will see what other files are able to be cached to save space. |
|
Thanks for the updates, @simrantan! |
I tested eos3ev6 and eos3ae6 on all three platforms, and they both ran successfully on eml_canonical. I also helped Riley investigate an issue by fetching a model on my laptop to see if the error he was receiving was a Mac-based one. We found that it persists on my OS, so it is in the model itself. I also thoroughly looked into a new model test error that occurred in actions when eos2rd8 was merged - partway through looking, I saw Riley message about the same error, and communicated with him about it, finding that I was looking in the wrong place for the error source. I then examined the error in more detail, did some research online, and wrote up my findings in issues. . |
I began working on eos8fth but found there was a fetch error that led me to discover the model might not be ready and I wanted to check on that before refactoring further. I also worked on refactoring eos5axz, which involved adding a main.py, moving functionality from service to main, and re-doing service.py. I spent a while making sure I was getting the format right and not messing up the functionality, but I am getting an error described here that I could advice on. I also tested eos76wn on CLI, Colab, and Docker and found it worked on CLI and Docker but had a fetch error in Colab. |
|
Hi @simrantan ! A few updates: |
|
I have continued work on eos5axz, and I am still receiving an error I find confusing about Nonetype objects. . I have also worked on eos8fth today, downloading and testing the model and checking for necessary files. I have some questions that I would love some guidance about before I move forward with incorporation here. |
|
Thanks for the update @simrantan Let's focus on completing the eos8fth, I've asked you a few questions about it in the issue |
I worked on eos8fth today, collecting information about the two different run_scripts and starting to incorporate the model by moving over the necessary files (including the model pkls). I have also changed the paths to reflect the new folders that information is found in from the original source code. . I am working on an input adapter, and sketched out some code for what an output adapter could look like. I also tested eos238c on CLI, Colab, and Docker. |
I had a meeting with Miquel and Zakia where they started helping with getting eos2rd8 set up locally so that I could build the image on my device and upload it to work around the actions size issues. I also tested eos1af5 on all platforms. I took some time to clean up forks of models I am no longer working on and removing codespaces I am not using. I started fresh on eos5axz, removing legacy files and making sure it works with "run", then submitting a PR. Finally, I worked on writing an input and output adapter for eos8fth. |
I tested eos6tg8 on CLI, Colab, and Docker and it worked great on all platforms. I finished writing the output adapter for eos8fth, and also wrapped up the last steps for incorporating the model but I ran into some bugs I am working on resolving (and will continue working on tomorrow). I have finished most of the steps for eos8fth, and once I get through these last bugs I should be able to finish testing it and submit it. |
Zakia reported an issue with eos5axz, which I looked into today and reported my findings in the issues thread as I wait for more details from Zakia. I also spent most of my time debugging eos8fth. I am working through some bugs but have gotten stuck on some that I have written about. Additionally, I spent some time testing a model for Riley since there was a potential Mac issue, then we worked on looking into possible solutions together that we can test further. |
I tested eos1579 on CLI, Colab, and Docker. I continued working on debugging eos8fth, working on the getting the remaining files into lfs. Zakia helped me look into the bug I was stuck on for eos8fth with the File error, and while I waited for her to try it on her CLI, I also started refactoring eos8451. I found a workflow error that I wrote up, and worked on debugging it. I fixed the issue, and I have refactored the model, but I have a question about the placement of main.py that I'd like clarification on before I consider it fully finished! Finally, I spent more time debugging eos8fth. With Zakias help, I have fixed the missing File bug, but I am still working on lfs files and a new error that has occurred. I plan to dig into eos8fth again tomorrow and hopefully resolve this issue. |
I spent most of today working on bug fixing for eos8451 and eos8fth. I managed to find a solution for my previous bug in eos8fth, and am working on testing to find the root of the error in eos8451. I unfortunately ran into a setback when I faced some laptop issues today (my computer shut down unexpectedly and would not turn back on), but managed to resolve it after calling the microsoft help desk. After the shutdown, I was advised to clean up storage, so I decided to do that for Ersilia as well, cleaning up any unnecessary resources. |
|
Hi @simrantan Please update the |
|
Hi, Sorry! I forgot to hit comment on my internship update for today. Also apologies for the lack of update on building the docker images - I have been working on creating an environment in docker but I haven't had success so I haven't had much progress to update on besides finding different missing libraries. I ran into some issues and wanted to possibly talk to Miquel and Zakia again to work through these problems, but since last week was busy I ended up focusing on eos8fth and other models in the meantime. (update below)
I spent some time working on eos8fth. Fixing the dependencies worked and I am able to start running the model. I've been working out some path issues as they pop up, but I think there are only a few remaining. I also had to get rid of their many print statements throughout the code that were printing to terminal, so I spent some time hunting those down and removing them. However, I am still having issues with lfs and tracking files on that model, which I have been working on for a long time. I am hoping to resolve Tuesday. but may need some input from anyone who has dealt with similar issues. I also am still working on trying to solve the path issue for eos8451 (the model takes a very long time to fetch, I have to wait a while to see the results of any new path I try). I found Riley is dealing with similar issues and we will tackle it together tomorrow. |
I have made a lot of progress on eos8fth and will be able to do final tests and wrap it up tonight or tomorrow depending on how long it takes to review file sizes for the lfs issue which I am still working on. I have also been working on debugging eos8451, and I have been using Febie's model eos85a3 as a guide since it is also a grover model. I am currently testing it, it just takes some time to fetch and will update with the success when it completes. I set up a meeting time with Miquel and Zakia to work on the local docker installation and started recording the process of commands I have been using to try and set up the environment (though it is not working, I am writing it down so that I can either refer back to these commands as a debugging tool or to provide alternative options) |
|
@GemmaTuron @miquelduranfrigola
As we discussed in the meeting, I am working on wrapping the models I have been working on and will be done with them soon! I would love to take on new models/testing soon so I can continue contributing. (pre-meeting update) I have almost finished eos8fth, but there is a dependency issue that I am unsure how to tackle since it deals with the version of docker ersilia uses and I'm unsure what should or shouldn't be changed for it to work. I also have fixed my prior issue with eos8451 |
I have been working on both of these issues today and reached points where I would love some outside input if available since I am dealing with things outside of the scope of the models themselves and do not want to mess up the ersilia model hub in any way. Eos8451 is dealing with an input issue, and eos8fth is having an empty output error (that is likely a package issue, since this module produces output just fine in a seperate conda environment using bash) |
Last friday, I met with Miquel and discussed the dockerhub issue. We went through the process, and found that the upload itself can only be completed by specific members of the Ersilia organization due to the permissions needed. I wrote out the steps we took to upload the file to Dockerhub and simplified to a general process that can be followed when encountering models that are too large for github actions to upload. I completed the draft of the writeup, and plan to add pictures and more details tomorrow. I also fixed eos8451 and submitted a PR. Inbetween, I helped Riley fetch and test eos5smc. Then, I worked on fixing eos8fth. |
|
thanks for the update @simrantan ! |
|
A comment on |
Worked on making the presentation for Wednesday's meeting. Tested eos8a5g on all platforms. Messaged Zakia and Hellen about rdkit, and tried looking for versions online. Have been testing different rdkit versions throughout the day (the model takes a while to fetch, so i have to wait some time after trying each version. |
|
8/2 and 8/3
Tested all the models I was assigned and addressed an issue in one of the models I was testing. I also was communicated with Zakia in the slack about the eos8fth rdkit issue |
|
8/4-8/6
I cleaned and refactored the assigned models (and also fixed the workflow error in eos6afz, which turned out to be an rdkit version issue). I have also been downloading different versions of rdkit and checking the number of descriptors. I have checked 5 versions but haven't found an exact match yet and am planning to test more tomorrow so I can get on to the next steps of checking if the descriptors are a match. |
|
Hi @simrantan IS model eos8fth ready? Please remember to move the tasks to In progress when you pick up new models.
You can get inspiration from this piece of code under the section #RDKIT Descriptors. HEre, we are selecting 200 specific descriptors |
|
Sorry! I will make sure to move the tasks to in progress in the future. Thank you for the files! I was trying Zakia's method of finding a version of rdkit that does work but it appears it will not be possible so I wrote some code to get the descriptors from the version the source code uses and have written a descriptors.py file based on the example provided! I had a question about how/where to use descriptors.py for it to prevent the issue from happening? Should I use it in the place of importing rdkit? Also, I moved the finetune folder to checkpoints and re-tested eos6fza and resubmitted a pr! |
|
For the descriptors, let's talk about this in today's meeting, I'll guide you through it! |
|
Hi @simrantan In addition to your tasks, please try out the test module developed by Riley and Febie and provide feedback in the Slack thread, specifying which model did you test, and if you have any further suggestions for improvement. Thanks! |
|
8/8-8/9
|
|
I had a flight scheduled for tonight that got cancelled last minute and I was moved to the next available one, which conflicts with tomorrows meeting so I will be unable to attend - I'll look at the presentations in the folder to catch up on the recent updates! |
|
Thanks for letting me know, safe flights. |
|
Hi @simrantan Let's focus on closng the open models eos1vms and eos8fth this week, thanks! |
|
Hi @simrantan ! It was great to work with you, thanks so much for your contributions to Ersilia, we hoped you learnt and enjoyed as much as we did! Please, remain engaged with the community and feel free to open any issues or contribute to open ones :) |
Summary
Hello,
This is a public issue for a virtual daily stand-up. We will use this to briefly share the tasks of the day and the challenges and advances made, so that we can ensure smooth support from the Ersilia mentors and alignment between daily tasks and overall internship goals.
Scope
Initiative 🐋
Objective(s)
Internship goals:
Team
Timeline
Before starting your work, line up a few tasks and short description. This should not take long. For example, it could be something like:
Wednesday 21st June
Documentation
No response
The text was updated successfully, but these errors were encountered: