|
| 1 | +import os |
| 2 | +from dataclasses import dataclass |
| 3 | + |
| 4 | +import torch |
| 5 | +from rydberggpt.models.rydberg_encoder_decoder import get_rydberg_graph_encoder_decoder |
| 6 | +from rydberggpt.models.utils import generate_prompt |
| 7 | +from rydberggpt.observables.rydberg_energy import ( |
| 8 | + get_rydberg_energy, |
| 9 | + get_staggered_magnetization, |
| 10 | + get_x_magnetization, |
| 11 | +) |
| 12 | +from rydberggpt.utils import create_config_from_yaml, load_yaml_file |
| 13 | +from rydberggpt.utils_ckpt import get_model_from_ckpt |
| 14 | +from torch_geometric.data import Batch |
| 15 | + |
| 16 | +device = "cuda" if torch.cuda.is_available() else "cpu" |
| 17 | + |
| 18 | +base_path = os.path.abspath(".") |
| 19 | +log_path = os.path.join(base_path, "models/ds_1/") |
| 20 | + |
| 21 | +yaml_dict = load_yaml_file(log_path, "hparams.yaml") |
| 22 | +config: dataclass = create_config_from_yaml(yaml_dict) |
| 23 | + |
| 24 | +model = get_model_from_ckpt( |
| 25 | + log_path, model=get_rydberg_graph_encoder_decoder(config), ckpt="best" |
| 26 | +) |
| 27 | +model.to(device=device) |
| 28 | +model.eval() # don't forget to set to eval mode |
| 29 | + |
| 30 | + |
| 31 | +def test_observables(): |
| 32 | + L = 5 |
| 33 | + delta = 1.0 |
| 34 | + omega = 1.0 |
| 35 | + beta = 64.0 |
| 36 | + Rb = 1.15 |
| 37 | + num_samples = 5 |
| 38 | + |
| 39 | + pyg_graph = generate_prompt( |
| 40 | + model_config=config, |
| 41 | + n_rows=L, |
| 42 | + n_cols=L, |
| 43 | + delta=delta, |
| 44 | + omega=omega, |
| 45 | + beta=beta, |
| 46 | + Rb=Rb, |
| 47 | + ) |
| 48 | + |
| 49 | + # duplicate the prompt for num_samples |
| 50 | + cond = [pyg_graph for _ in range(num_samples)] |
| 51 | + cond = Batch.from_data_list(cond) |
| 52 | + |
| 53 | + samples = model.get_samples( |
| 54 | + batch_size=len(cond), cond=cond, num_atoms=L**2, fmt_onehot=False |
| 55 | + ) |
| 56 | + |
| 57 | + energy = get_rydberg_energy(model, samples, cond=pyg_graph, device=device) |
| 58 | + print(energy.mean() / L**2) |
| 59 | + |
| 60 | + staggered_magnetization = get_staggered_magnetization(samples, L, L, device=device) |
| 61 | + print(staggered_magnetization.mean() / L**2) |
| 62 | + |
| 63 | + x_magnetization = get_x_magnetization(model, samples, cond=pyg_graph, device=device) |
| 64 | + print(x_magnetization.mean() / L**2) |
| 65 | + assert True |
| 66 | + |
| 67 | + |
| 68 | +test_observables() |
0 commit comments