Comments (3)
Ok, I believe the probelm is id2emb[idx]
? I used Michael's way of dealing with loading embeddings which is not casting them to a type until this very moment, we could just cast the embeddings to np.arrays when they are loaded. I'll see how far I get to fix this in the inference PR
from biotrainer.
Yes, that (concatenating all samples to one long numpy array) will work for per_protein tasks (a list of 10 elements each holding a 1024-d array would became a 10x1024 matrix) but for per_residue tasks, you would need to pad all sequences in your set to the longest sequence. This is usually not the desired behavior which is why I kept this list-structure. I also do not see an easy way around this.
from biotrainer.
The warning disappeared and the performance seems to be fine, so I will close this issue for now.
from biotrainer.
Related Issues (20)
- Per-residue predictions differ between batches and single input embeddings
- Implement embeddings calculation "on the fly"
- Split annotations are more complicated than necessary HOT 1
- Add random comparison baseline
- Config - Embeddings - Targets pipeline is inefficient HOT 1
- Class weights need to be sent to the device when use_class_weights=True
- Support multiple hyperparameters for hold_out cross validation
- [ppi] Concat does not work for 0-dimensional (scalar) tensors
- Adding BERT model and protocol = transformer encoder model + masked language modeling (MLM)
- [ppi] Interaction mode not compatible with all protocols yet
- Inferencer does not set correct device
- Create CI and docker image HOT 1
- Create tutorial how-to use a custom embedder
- Update dependencies
- Value Error when Early stopping was engaged HOT 2
- Do a sanity check of the inferencer module
- :woman_scientist: Add LoRA layers to fine-tune protein language models for embeddings calculation
- Look into AI Code suggestions
- elapsed_time_total is misleading
- Improve embeddings saving strategy HOT 1
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 biotrainer.