Update README.md

README.md

@@ -6,6 +6,16 @@ tags: []
 # Model Card for Model ID
 
 <!-- Provide a quick summary of what the model is/does. -->
+Pretrained Foundation Neural Quantum State on the Ising in disordered transverse field model on a chain with \\(L=32\\) sites.
+Several values of the external field are available (check the different revisions).
+
+The model has been trained on \\(R=2000\\) different disorder realization, using a total batch size of \\(M=10000\\) samples.
+The computation has been distributed over 4 A100-64GB GPUs for about two hours. 
+
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model. In particular, we sample the model for a fixed disordered realization using NetKet.
 
 ```python
 from functools import partial
@@ -77,3 +87,24 @@ for _ in range(10):
     print("Mean: ", E.mean.real / lattice.n_nodes, "\t time=", time.time()-start)
 ```
 
+The time per sweep is 0.5s, evaluated on a single A100-40GB GPU.
+
+### Extract hidden representation
+
+The hidden representation associated to the input batch of configurations can be extracted as:
+
+```python
+wf = FlaxAutoModel.from_pretrained("nqs-models/ising_disorder_fnqs", trust_remote_code=True, return_z=True)
+
+z = wf(wf.params, samples, h)
+```
+
+#### Training Hyperparameters
+
+Number of layers: 6
+Embedding dimension: 72   
+Hidden dimension: 288  
+Number of heads: 12  
+Patch size: 4
+
+Total number of parameters: 326124