Dual-unitary recurrent state block / results.json

Mechanism failed

Raw ⬇ ZIP
 1{
 2  "device": "cpu",
 3  "math": {
 4    "Uplus_orth_error": 1.414555065850144e-15,
 5    "Uminus_orth_error": 1.414555065850144e-15,
 6    "block_sign_construction_error": 0.0,
 7    "difference_structure_error": 0.0,
 8    "power_difference_identity_error": 1.1216137658185826e-15,
 9    "max_norm_error_over_powers": 4.440892098500626e-15
10  },
11  "experiment": {
12    "tanh_rnn": {
13      "final_train_loss": 0.6932020115852356,
14      "test_loss": 0.6930223107337952,
15      "accuracy": 0.5121231079101562,
16      "hidden_norm_last": 0.5228166580200195,
17      "hidden_norm_max": 1.4647603034973145,
18      "input_grad_first": 1.2615543332206158e-10,
19      "input_grad_last": 0.0003851530491374433
20    },
21    "single_unitary": {
22      "final_train_loss": 0.693144121170044,
23      "test_loss": 0.6932753920555115,
24      "accuracy": 0.5005340576171875,
25      "hidden_norm_last": 3.9796600341796875,
26      "hidden_norm_max": 3.9796600341796875,
27      "input_grad_first": 0.0008569805650040507,
28      "input_grad_last": 0.00011255657591391355
29    },
30    "dual_unitary": {
31      "final_train_loss": 0.6931803965568543,
32      "test_loss": 0.693268358707428,
33      "accuracy": 0.5,
34      "hidden_norm_last": 3.201139211654663,
35      "hidden_norm_max": 3.201139211654663,
36      "input_grad_first": 0.0008913782075978816,
37      "input_grad_last": 2.9597111279144883e-05
38    }
39  },
40  "elapsed_sec": 62.8214898109436
41}