akore commited on
Commit
1164eae
1 Parent(s): bd07045

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +7 -9
README.md CHANGED
@@ -35,16 +35,17 @@ is not validated, it was primarily used a pre-training task.
35
 
36
  ### How to use
37
 
38
- You can use this model directly by loading via the Structure2EnergyandForces task:
39
 
40
  ```python
41
- >>> from transformers import AutoModel
42
- ```
 
43
 
44
- Here is how to use this model to get the features of a given atomistic graph in PyTorch:
45
 
46
- ```python
47
- from transformers import AutoModel
48
  ```
49
 
50
 
@@ -54,9 +55,6 @@ AtomFormer is trained on an aggregated S2EF dataset from multiple sources such a
54
  with structures and energies/forces for pre-training. The pre-training data includes total energies and formation
55
  energies but trains using formation energy (which isn't included for OC22, indicated by "has_formation_energy" column).
56
 
57
- ## Training procedure
58
-
59
-
60
 
61
  ### Preprocessing
62
 
 
35
 
36
  ### How to use
37
 
38
+ Here is how to use the model to extract features from the pre-trained backbone:
39
 
40
  ```python
41
+ import torch
42
+ from transformers import AutoModel
43
+ model = AutoModel.from_pretrained("vector-institute/atomformer-base", trust_remote_code=True)
44
 
45
+ input_ids, coords, attn_mask = torch.randint(0, 100, (1, 10)), torch.randn(1, 10, 3), torch.ones(1, 10)
46
 
47
+ output = model(input_ids, coords=coords, attention_mask=attention_mask)
48
+ output[0].shape # (torch.Size([1, 10, 768])
49
  ```
50
 
51
 
 
55
  with structures and energies/forces for pre-training. The pre-training data includes total energies and formation
56
  energies but trains using formation energy (which isn't included for OC22, indicated by "has_formation_energy" column).
57
 
 
 
 
58
 
59
  ### Preprocessing
60