Comments (8)
I'm not familiar with this 1280-element embedding. Could you elaborate? Are you talking about what DeepMind calls the "single" embedding?
from openfold.
I'm not familiar with this 1280-element embedding. Could you elaborate? Are you talking about what DeepMind calls the "single" embedding?
Thanks for your quick response. I'm talking about the Evolutionary Scale Modeling(ESM) from FACEBOOK. for each protein sequence they output just one single embedding that represents the whole protein sequence in tabular form #aminoacids*1280. this embedding has information about the protein sequence and I want to extract this from the inner layers of Openfold. I'm looking for sth like this embedding in the Openfold. could you give me the details about this single embedding from DeepMind? maybe this is the one that I'm looking for.
from openfold.
It's one of the outputs of the model (I think it's called "single" in the output dictionary), and is derived from the first row of the MSA embedding that comes out of the evoformer. It has shape [N, 384].
from openfold.
Whoops---didn't mean to close. If you want the single sequence embedding from different points in the model, you could look at its value during different recycling iterations, for example.
from openfold.
could you elaborate on the "different points in the model". also I have this question, if I just run the code from the Colab version that you provide, is this information available in the zip file that it generates?
from openfold.
No, it's not in the output zip by default. You'll have to slightly modify the Colab to extract more stuff from the output dictionary.
As for "different points in the model:" you asked about viewing the single embedding in the "inner layers" of OpenFold. I assumed that meant you want to observe how the single embedding evolves over time. One way to do that would be to inspect its value after each recycling iteration.
from openfold.
No, it's not in the output zip by default. You'll have to slightly modify the Colab to extract more stuff from the output dictionary.
As for "different points in the model:" you asked about viewing the single embedding in the "inner layers" of OpenFold. I assumed that meant you want to observe how the single embedding evolves over time. One way to do that would be to inspect its value after each recycling iteration.
Thanks for the explanation. could you please also give me some instructions on how to extract these single embeddings from the Colab version or even from the original code that I could follow?
That would be a great help and I really appreciate it.
from openfold.
The model outputs a dictionary. In the Colab, only the structure and pLDDT scores are currently retrieved from it. Getting the single embedding is as simple as retrieving "single" from the dictionary.
from openfold.
Related Issues (20)
- Inference error using precomputed alignments
- Pickling error in Docker? HOT 1
- Is the pdb_mmcif.zip from OpenProteinSet on AWS suitable for RODA training data? HOT 1
- Finetune from AF_Multimer parameters HOT 7
- Unable to create conda environment HOT 5
- I notice there is only instructions for a Linux Install, can I run this on my Windows laptop? HOT 1
- Frequently failed in training. HOT 2
- Rigid.from_3_points comment HOT 1
- Colab broken by version skew HOT 1
- Docker build broken HOT 4
- Question: Can the geometry module and rigid_utils be converted to each other?
- Alignment error during inference HOT 1
- ModuleNotFoundError: No module named 'attn_core_inplace_cuda' HOT 3
- Enable Dropout in inference HOT 1
- Questions about the meaning of folder naming conventions in OpenProteinSet HOT 1
- ModuleNotFoundError: No module named 'attn_core_inplace_cuda' HOT 1
- RuntimeError: Error building extension 'evoformer_attn' HOT 1
- Multimer predicting a homomer HOT 2
- Docker container AttributeError: Did you mean: 'linear_a_p'? when trying to run multimer inference
- Unable to install OpenFold in Google Colaboratory
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from openfold.