Weighted Resolvent-Equivariant Attention / results.json
Unverified
1{
2 "seed": 2327,
3 "math": {
4 "weighted_G": [
5 [
6 0.0,
7 0.0,
8 0.0,
9 0.0,
10 0.0,
11 1.0
12 ],
13 [
14 0.0,
15 0.0,
16 0.0,
17 0.0,
18 1.7,
19 0.0
20 ],
21 [
22 0.0,
23 0.0,
24 0.0,
25 0.8,
26 0.0,
27 0.0
28 ],
29 [
30 0.0,
31 0.0,
32 1.3,
33 0.0,
34 0.0,
35 0.0
36 ],
37 [
38 0.0,
39 2.0,
40 0.0,
41 0.0,
42 0.0,
43 0.0
44 ],
45 [
46 0.6,
47 0.0,
48 0.0,
49 0.0,
50 0.0,
51 0.0
52 ]
53 ],
54 "prediction_1_linear_perturbation": {
55 "eps": [
56 0.0,
57 0.001,
58 0.003,
59 0.01,
60 0.03,
61 0.1,
62 0.3
63 ],
64 "observed_commutators": [
65 0.0,
66 0.0019490096902629438,
67 0.005847029070788675,
68 0.019490096902629192,
69 0.058470290707887486,
70 0.19490096902629167,
71 0.584702907078875
72 ],
73 "predicted_slope_norm_[G,E]": 1.9490096902629164,
74 "fit_slope": 1.9490096902629166,
75 "max_relative_deviation": 1.898781434441184e-16
76 },
77 "prediction_2_exact_multi_step": {
78 "power_commutator_norms_t0_to_8": [
79 0.0,
80 0.0,
81 0.0,
82 0.0,
83 3.925231146709438e-17,
84 3.260541871072589e-16,
85 6.341404989338363e-16,
86 6.329245045284193e-16,
87 6.329245045284193e-16
88 ],
89 "resolvent_commutator_norm_gamma_0.37": 5.042050700450315e-16
90 },
91 "prediction_3_damping_resolvent": {
92 "gamma": [
93 0.0,
94 0.1,
95 0.25,
96 0.5,
97 0.7,
98 0.85,
99 0.95,
100 1.0
101 ],
102 "observed_norms": [
103 68.58527359241329,
104 114.5429542276119,
105 2.2774615710080095,
106 0.3265899854079163,
107 0.0971897699096757,
108 0.032600702382032394,
109 0.008656103548493213,
110 0.0
111 ],
112 "normalized_to_gamma_0": [
113 1.0,
114 1.6700808822067938,
115 0.033206276678904086,
116 0.004761809180040113,
117 0.0014170646965302267,
118 0.00047533093730544793,
119 0.00012620936091812466,
120 0.0
121 ],
122 "resolvent_identity_relative_errors": [
123 1.0397771809966086e-15,
124 3.3531591390592183e-15,
125 1.5671720216967223e-15,
126 2.844462506732615e-15,
127 8.54015480364294e-15,
128 1.4565979322802336e-14,
129 1.109860327670625e-13,
130 0.0
131 ],
132 "near_gamma_1_fit_slope_vs_(1-gamma)": 0.21291642138918074,
133 "gamma_1_norm": 0.0
134 }
135 },
136 "attention_mini_experiment": {
137 "0.0": {
138 "seeds": [
139 {
140 "accuracy": 0.7142857313156128,
141 "val_loss": 0.5585198402404785,
142 "commutator_rms": 3.419782192389815,
143 "device": "cuda",
144 "train_loss": 0.3901596665382385
145 },
146 {
147 "accuracy": 0.6845238208770752,
148 "val_loss": 0.7168405055999756,
149 "commutator_rms": 3.3688179521321975,
150 "device": "cuda",
151 "train_loss": 0.4374329149723053
152 }
153 ],
154 "mean_accuracy": 0.699404776096344,
155 "mean_val_loss": 0.637680172920227,
156 "mean_commutator_rms": 3.3943000722610064
157 },
158 "1.0": {
159 "seeds": [
160 {
161 "accuracy": 0.7559524178504944,
162 "val_loss": 0.4704103469848633,
163 "commutator_rms": 0.3829123576741427,
164 "device": "cuda",
165 "train_loss": 0.4502479135990143
166 },
167 {
168 "accuracy": 0.7142857313156128,
169 "val_loss": 0.565798819065094,
170 "commutator_rms": 0.4035790097316604,
171 "device": "cuda",
172 "train_loss": 0.498306006193161
173 }
174 ],
175 "mean_accuracy": 0.7351190745830536,
176 "mean_val_loss": 0.5181045830249786,
177 "mean_commutator_rms": 0.3932456837029016
178 }
179 }
180}