Comments (12)
Hey @shenoynikhil, thank you for email.
This is entirely possible. Obviously you would need some of the arguments of run_train to create the model. You can look at the model test to check how it is done: test_models.py
from mace.
This looks great! Thank you so much. I'm closing the issue.
from mace.
I was trying to use AtomicData object with pytorch_geometric DataLoader. Because its not torch_geometric.data.Data
, the Batch.from_data_list([<list of AtomicData>])
does not work. Do you have an idea of what could be done to fix this?
I saw that you guys have copied the dataloader, collate stuff in mace.tools.torch_geometric
and use that for dataloader but I could not figure out the difference that allows this to be loaded.
from mace.
Could you give a little more detail of the problem? I recommend that you import the data loader from mace, not from torch geometric.
from mace.
Currently, if I just install pytorch_geometric and use their Batch.from_data_list([<list of AtomicData>])
it does not run. I was able to get it to run by copying your batch.py
.
I do not want to use your torch_geometric since I am benchmarking this with other networks and other molecular graph datasets and want to use the latest pyg
.
from mace.
We use a lightweight version of Pytorch geometric to ease the maintenance and installation. Pytorch geometric is notoriously hard to install due to Cuda extensions. The mace models now take as input a dictionary of tensors. To make your code compatible, create the dictionaries with the correct entries. You should be able to do that from any initial Python object.
from mace.
Okay, is there any documentation (positions
, edge_index
, node_attrs
etc) on what keys would be required, or is that something for which I would need to go through the code?
from mace.
Currently, it is not documented, but very easy to check by looking at the same test test_models.py and printing the dictionary.
from mace.
I want to add that training mace with the provided training script is highly encouraged, as many optimization details are crucial to mace performance.
from mace.
The thing is, I want to test training this model on QM datasets that only have energies (also not present in xyz format). And I believe if the model is not tied to your repo and data utils, it might be easier to adopt. Kind of like how dimenet++ is present in pyg.
from mace.
Only energies are also supported currently. I just wanted to warn you that training these ML force fields can be tricky, and I recommend you at least train on this repo with the default optimisation procedure to have a reference performance to compare to.
I have made another repo mace-layer that lets your import a mace layer and stack it however you like with very standard and documented inputs also.
from mace.
Only energies are also supported currently. I just wanted to warn you that training these ML force fields can be tricky, and I recommend you at least train on this repo with the default optimisation procedure to have a reference performance to compare to
Good Option. I'll make sure to do that.
I have made another repo mace-layer that lets your import a mace layer and stack it however you like with very standard and documented inputs also.
Great, will check it out.
from mace.
Related Issues (20)
- Error reading 'extended' xyz format files HOT 23
- What are the node energies? HOT 2
- Training flag not working HOT 2
- Dipole fitting on multi-GPU
- Is 0.694 timesteps/s a normal speed?
- select pretrained configs for multihead replay with or without replacement? HOT 2
- adding cmd argument for specifying wandb directory HOT 2
- CUDA Out of Memory with GCMC
- Distributed Training: invalid device ordinal
- Missing dependancies in `multi-head-interface` HOT 1
- Stress / num_atoms in loss? HOT 3
- stress errors reported during fitting and at end should be consistent HOT 4
- during-training valid loss is wrong HOT 11
- np.random.choice for selecting pre-fit subset for multihead fine tune doesn't accept a list of Atoms
- multi-head-finetune validation batches not deterministic HOT 3
- installation HOT 6
- How to Convert VASP Output Files to MACE XYZ Format for Training and Testing Sets
- How to use multi-training without slurm system? HOT 11
- Problem in Compiling Lammps for GPU version HOT 3
- Error in online data loading for large datasets 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 mace.