Critical-Batch Momentum Scaling / results.json
Mechanism failed
1{
2 "math_check": [
3 {
4 "kind": "polyak",
5 "rho": 0.0,
6 "critical": 1.9999999999999998,
7 "ratio": 0.5263157894736841,
8 "predicted_uncapped_ratio": 1.0
9 },
10 {
11 "kind": "polyak",
12 "rho": 0.5,
13 "critical": 2.9999999999999996,
14 "ratio": 0.7894736842105261,
15 "predicted_uncapped_ratio": 1.0
16 },
17 {
18 "kind": "polyak",
19 "rho": 0.9,
20 "critical": 3.8000000000000003,
21 "ratio": 1.0,
22 "predicted_uncapped_ratio": 1.0
23 },
24 {
25 "kind": "polyak",
26 "rho": 0.95,
27 "critical": 3.899999999999999,
28 "ratio": 1.0263157894736838,
29 "predicted_uncapped_ratio": 0.5000000000000006
30 },
31 {
32 "kind": "nesterov",
33 "rho": 0.0,
34 "critical": 1.9999999999999998,
35 "ratio": 1.4736842105263157,
36 "predicted_uncapped_ratio": 1.0
37 },
38 {
39 "kind": "nesterov",
40 "rho": 0.5,
41 "critical": 1.4999999999999998,
42 "ratio": 1.1052631578947367,
43 "predicted_uncapped_ratio": 1.0
44 },
45 {
46 "kind": "nesterov",
47 "rho": 0.9,
48 "critical": 1.357142857142857,
49 "ratio": 1.0,
50 "predicted_uncapped_ratio": 1.0
51 },
52 {
53 "kind": "nesterov",
54 "rho": 0.95,
55 "critical": 1.3448275862068964,
56 "ratio": 0.9909255898366606,
57 "predicted_uncapped_ratio": 0.5000000000000006
58 }
59 ],
60 "controller_runs": [
61 {
62 "kind": "polyak",
63 "B": 1,
64 "eta_controller": 0.025,
65 "fixed_stable": true,
66 "controller_stable": true,
67 "fixed_final_loss": 1.3158477888406337,
68 "controller_final_loss": 0.6304000260049999,
69 "controller_update": 0.1557288472657435
70 },
71 {
72 "kind": "polyak",
73 "B": 2,
74 "eta_controller": 0.05,
75 "fixed_stable": true,
76 "controller_stable": true,
77 "fixed_final_loss": 0.6488216988289215,
78 "controller_final_loss": 0.6488216988289215,
79 "controller_update": 0.2168853635564954
80 },
81 {
82 "kind": "polyak",
83 "B": 4,
84 "eta_controller": 0.05,
85 "fixed_stable": true,
86 "controller_stable": true,
87 "fixed_final_loss": 0.09127614731111712,
88 "controller_final_loss": 0.09127614731111712,
89 "controller_update": 0.16405308143917
90 },
91 {
92 "kind": "polyak",
93 "B": 8,
94 "eta_controller": 0.05,
95 "fixed_stable": true,
96 "controller_stable": true,
97 "fixed_final_loss": 0.09854773474795911,
98 "controller_final_loss": 0.09854773474795911,
99 "controller_update": 0.10463650049094794
100 },
101 {
102 "kind": "polyak",
103 "B": 16,
104 "eta_controller": 0.05,
105 "fixed_stable": true,
106 "controller_stable": true,
107 "fixed_final_loss": 0.027525625328528647,
108 "controller_final_loss": 0.027525625328528647,
109 "controller_update": 0.09590454539789173
110 },
111 {
112 "kind": "nesterov",
113 "B": 1,
114 "eta_controller": 0.025,
115 "fixed_stable": true,
116 "controller_stable": true,
117 "fixed_final_loss": 0.9043768015759052,
118 "controller_final_loss": 0.5101583218552774,
119 "controller_update": 0.1446629868284835
120 },
121 {
122 "kind": "nesterov",
123 "B": 2,
124 "eta_controller": 0.05,
125 "fixed_stable": true,
126 "controller_stable": true,
127 "fixed_final_loss": 0.4037268043042466,
128 "controller_final_loss": 0.4037268043042466,
129 "controller_update": 0.17898134730227575
130 },
131 {
132 "kind": "nesterov",
133 "B": 4,
134 "eta_controller": 0.05,
135 "fixed_stable": true,
136 "controller_stable": true,
137 "fixed_final_loss": 0.0768287269915019,
138 "controller_final_loss": 0.0768287269915019,
139 "controller_update": 0.13843437164181138
140 },
141 {
142 "kind": "nesterov",
143 "B": 8,
144 "eta_controller": 0.05,
145 "fixed_stable": true,
146 "controller_stable": true,
147 "fixed_final_loss": 0.0641408837970391,
148 "controller_final_loss": 0.0641408837970391,
149 "controller_update": 0.08409418140893751
150 },
151 {
152 "kind": "nesterov",
153 "B": 16,
154 "eta_controller": 0.05,
155 "fixed_stable": true,
156 "controller_stable": true,
157 "fixed_final_loss": 0.017482954889978204,
158 "controller_final_loss": 0.017482954889978204,
159 "controller_update": 0.07796698336954941
160 }
161 ],
162 "stability_scan": [
163 {
164 "kind": "polyak",
165 "B": 1,
166 "empirical_max_eta": 2.0,
167 "deterministic_quadratic_eta": 3.8000000000000003
168 },
169 {
170 "kind": "polyak",
171 "B": 2,
172 "empirical_max_eta": 2.0,
173 "deterministic_quadratic_eta": 3.8000000000000003
174 },
175 {
176 "kind": "polyak",
177 "B": 4,
178 "empirical_max_eta": 2.0,
179 "deterministic_quadratic_eta": 3.8000000000000003
180 },
181 {
182 "kind": "polyak",
183 "B": 8,
184 "empirical_max_eta": 2.0,
185 "deterministic_quadratic_eta": 3.8000000000000003
186 },
187 {
188 "kind": "polyak",
189 "B": 16,
190 "empirical_max_eta": 2.0,
191 "deterministic_quadratic_eta": 3.8000000000000003
192 },
193 {
194 "kind": "nesterov",
195 "B": 1,
196 "empirical_max_eta": 1.2487759931307252,
197 "deterministic_quadratic_eta": 1.357142857142857
198 },
199 {
200 "kind": "nesterov",
201 "B": 2,
202 "empirical_max_eta": 1.2487759931307252,
203 "deterministic_quadratic_eta": 1.357142857142857
204 },
205 {
206 "kind": "nesterov",
207 "B": 4,
208 "empirical_max_eta": 1.2487759931307252,
209 "deterministic_quadratic_eta": 1.357142857142857
210 },
211 {
212 "kind": "nesterov",
213 "B": 8,
214 "empirical_max_eta": 1.2487759931307252,
215 "deterministic_quadratic_eta": 1.357142857142857
216 },
217 {
218 "kind": "nesterov",
219 "B": 16,
220 "empirical_max_eta": 1.2487759931307252,
221 "deterministic_quadratic_eta": 1.357142857142857
222 }
223 ],
224 "config": {
225 "eta0": 0.05,
226 "rho0": 0.9,
227 "beta": 1.5,
228 "safety": 0.5
229 }
230}