layer_type | N | M | Q | alpha | D | alpha-hat | log_SN | rand_D | num_traps | stable_rank | rank_loss | |
---|---|---|---|---|---|---|---|---|---|---|---|---|
layer_id | ||||||||||||
2 | EMBEDDING | 250002 | 768 | 325.52 | 2.37 | 0.04 | 14.78 | 6.25 | 0.38 | 1 | 4.89 | 0.00 |
3 | EMBEDDING | 768 | 514 | 1.49 | 2.53 | 0.06 | 8.10 | 3.20 | 0.72 | 1 | 1.94 | 1.00 |
12 | DENSE | 768 | 768 | 1.00 | 3.56 | 0.05 | 7.43 | 2.09 | 0.10 | 0 | 67.05 | 1.00 |
13 | DENSE | 768 | 768 | 1.00 | 7.72 | 0.07 | 17.98 | 2.33 | 0.24 | 0 | 41.72 | 1.00 |
14 | DENSE | 768 | 768 | 1.00 | 5.20 | 0.06 | 6.41 | 1.23 | 0.18 | 0 | 46.94 | 2.00 |
17 | DENSE | 768 | 768 | 1.00 | 4.03 | 0.06 | 6.75 | 1.67 | 0.17 | 1 | 16.76 | 2.00 |
21 | DENSE | 3072 | 768 | 4.00 | 2.53 | 0.04 | 7.94 | 3.14 | 0.29 | 1 | 7.66 | 0.00 |
24 | DENSE | 3072 | 768 | 4.00 | 6.28 | 0.06 | 11.59 | 1.85 | 0.08 | 0 | 72.74 | 0.00 |
30 | DENSE | 768 | 768 | 1.00 | 2.48 | 0.05 | 5.66 | 2.28 | 0.18 | 0 | 25.73 | 2.00 |
31 | DENSE | 768 | 768 | 1.00 | 2.55 | 0.08 | 5.19 | 2.04 | 0.19 | 0 | 44.30 | 2.00 |
32 | DENSE | 768 | 768 | 1.00 | 4.68 | 0.08 | 5.09 | 1.09 | 0.15 | 0 | 68.14 | 1.00 |
35 | DENSE | 768 | 768 | 1.00 | 4.63 | 0.07 | 4.63 | 1.00 | 0.11 | 0 | 68.76 | 2.00 |
39 | DENSE | 3072 | 768 | 4.00 | 2.62 | 0.03 | 8.08 | 3.09 | 0.20 | 1 | 8.35 | 0.00 |
42 | DENSE | 3072 | 768 | 4.00 | 5.09 | 0.02 | 9.01 | 1.77 | 0.11 | 0 | 79.82 | 0.00 |
48 | DENSE | 768 | 768 | 1.00 | 2.41 | 0.05 | 4.73 | 1.96 | 0.19 | 0 | 36.84 | 2.00 |
49 | DENSE | 768 | 768 | 1.00 | 2.63 | 0.05 | 5.22 | 1.98 | 0.19 | 0 | 35.11 | 2.00 |
50 | DENSE | 768 | 768 | 1.00 | 4.19 | 0.08 | 4.37 | 1.04 | 0.12 | 0 | 82.45 | 1.00 |
53 | DENSE | 768 | 768 | 1.00 | 3.31 | 0.04 | 4.46 | 1.35 | 0.16 | 0 | 30.99 | 3.00 |
57 | DENSE | 3072 | 768 | 4.00 | 2.67 | 0.03 | 8.09 | 3.03 | 0.19 | 1 | 8.81 | 0.00 |
60 | DENSE | 3072 | 768 | 4.00 | 3.20 | 0.06 | 5.79 | 1.81 | 0.17 | 0 | 87.94 | 0.00 |
66 | DENSE | 768 | 768 | 1.00 | 3.82 | 0.09 | 6.10 | 1.60 | 0.15 | 0 | 67.36 | 2.00 |
67 | DENSE | 768 | 768 | 1.00 | 3.69 | 0.07 | 5.75 | 1.56 | 0.12 | 0 | 74.83 | 2.00 |
68 | DENSE | 768 | 768 | 1.00 | 4.53 | 0.13 | 5.24 | 1.16 | 0.13 | 0 | 91.98 | 1.00 |
71 | DENSE | 768 | 768 | 1.00 | 2.43 | 0.06 | 3.42 | 1.41 | 0.20 | 0 | 45.03 | 1.00 |
75 | DENSE | 3072 | 768 | 4.00 | 3.22 | 0.03 | 10.09 | 3.14 | 0.18 | 1 | 7.80 | 0.00 |
78 | DENSE | 3072 | 768 | 4.00 | 4.77 | 0.03 | 10.40 | 2.18 | 0.17 | 0 | 47.94 | 0.00 |
84 | DENSE | 768 | 768 | 1.00 | 3.09 | 0.08 | 5.22 | 1.69 | 0.13 | 0 | 60.71 | 1.00 |
85 | DENSE | 768 | 768 | 1.00 | 3.23 | 0.07 | 5.44 | 1.68 | 0.13 | 0 | 64.18 | 1.00 |
86 | DENSE | 768 | 768 | 1.00 | 4.81 | 0.12 | 5.93 | 1.23 | 0.14 | 0 | 85.65 | 0.00 |
89 | DENSE | 768 | 768 | 1.00 | 2.03 | 0.08 | 3.59 | 1.77 | 0.25 | 0 | 33.71 | 3.00 |
93 | DENSE | 3072 | 768 | 4.00 | 3.59 | 0.04 | 11.75 | 3.27 | 0.20 | 1 | 6.65 | 0.00 |
96 | DENSE | 3072 | 768 | 4.00 | 4.56 | 0.03 | 10.96 | 2.40 | 0.20 | 0 | 32.96 | 0.00 |
102 | DENSE | 768 | 768 | 1.00 | 3.86 | 0.12 | 5.51 | 1.43 | 0.13 | 0 | 90.39 | 3.00 |
103 | DENSE | 768 | 768 | 1.00 | 3.28 | 0.06 | 5.34 | 1.62 | 0.15 | 0 | 60.23 | 4.00 |
104 | DENSE | 768 | 768 | 1.00 | 8.42 | 0.09 | 10.76 | 1.28 | 0.12 | 0 | 91.56 | 1.00 |
107 | DENSE | 768 | 768 | 1.00 | 1.89 | 0.12 | 3.00 | 1.59 | 0.23 | 0 | 56.04 | 2.00 |
111 | DENSE | 3072 | 768 | 4.00 | 3.75 | 0.04 | 12.24 | 3.27 | 0.20 | 1 | 6.31 | 0.00 |
114 | DENSE | 3072 | 768 | 4.00 | 4.04 | 0.03 | 9.77 | 2.42 | 0.24 | 0 | 29.25 | 0.00 |
120 | DENSE | 768 | 768 | 1.00 | 3.44 | 0.09 | 6.67 | 1.94 | 0.14 | 0 | 34.16 | 3.00 |
121 | DENSE | 768 | 768 | 1.00 | 3.24 | 0.10 | 5.06 | 1.56 | 0.14 | 0 | 83.46 | 4.00 |
122 | DENSE | 768 | 768 | 1.00 | 6.07 | 0.08 | 8.43 | 1.39 | 0.15 | 0 | 59.55 | 2.00 |
125 | DENSE | 768 | 768 | 1.00 | 1.96 | 0.10 | 3.20 | 1.64 | 0.25 | 0 | 45.92 | 2.00 |
129 | DENSE | 3072 | 768 | 4.00 | 3.62 | 0.05 | 11.68 | 3.23 | 0.20 | 1 | 6.47 | 0.00 |
132 | DENSE | 3072 | 768 | 4.00 | 3.56 | 0.03 | 9.40 | 2.64 | 0.26 | 0 | 17.25 | 0.00 |
138 | DENSE | 768 | 768 | 1.00 | 3.83 | 0.06 | 6.72 | 1.75 | 0.15 | 0 | 47.59 | 3.00 |
139 | DENSE | 768 | 768 | 1.00 | 3.40 | 0.09 | 5.41 | 1.59 | 0.17 | 0 | 71.46 | 4.00 |
140 | DENSE | 768 | 768 | 1.00 | 10.23 | 0.14 | 11.71 | 1.14 | 0.10 | 0 | 108.78 | 2.00 |
143 | DENSE | 768 | 768 | 1.00 | 2.10 | 0.08 | 3.48 | 1.66 | 0.20 | 0 | 37.25 | 2.00 |
147 | DENSE | 3072 | 768 | 4.00 | 3.58 | 0.05 | 11.24 | 3.14 | 0.19 | 1 | 7.18 | 0.00 |
150 | DENSE | 3072 | 768 | 4.00 | 3.77 | 0.03 | 9.70 | 2.57 | 0.25 | 0 | 18.85 | 0.00 |
156 | DENSE | 768 | 768 | 1.00 | 5.46 | 0.06 | 10.10 | 1.85 | 0.17 | 0 | 40.32 | 3.00 |
157 | DENSE | 768 | 768 | 1.00 | 3.90 | 0.13 | 6.02 | 1.54 | 0.17 | 0 | 82.93 | 4.00 |
158 | DENSE | 768 | 768 | 1.00 | 7.15 | 0.13 | 8.04 | 1.13 | 0.09 | 0 | 114.92 | 4.00 |
161 | DENSE | 768 | 768 | 1.00 | 1.90 | 0.07 | 3.25 | 1.71 | 0.28 | 0 | 38.54 | 1.00 |
165 | DENSE | 3072 | 768 | 4.00 | 3.61 | 0.04 | 11.02 | 3.06 | 0.19 | 1 | 8.17 | 0.00 |
168 | DENSE | 3072 | 768 | 4.00 | 3.61 | 0.02 | 8.90 | 2.47 | 0.23 | 0 | 24.46 | 0.00 |
174 | DENSE | 768 | 768 | 1.00 | 4.61 | 0.07 | 9.69 | 2.10 | 0.16 | 0 | 23.75 | 3.00 |
175 | DENSE | 768 | 768 | 1.00 | 3.82 | 0.09 | 6.17 | 1.62 | 0.16 | 0 | 72.21 | 4.00 |
176 | DENSE | 768 | 768 | 1.00 | 7.19 | 0.05 | 7.97 | 1.11 | 0.07 | 0 | 94.91 | 3.00 |
179 | DENSE | 768 | 768 | 1.00 | 4.00 | 0.07 | 6.61 | 1.65 | 0.15 | 0 | 26.88 | 1.00 |
183 | DENSE | 3072 | 768 | 4.00 | 3.54 | 0.05 | 10.49 | 2.96 | 0.17 | 1 | 8.60 | 0.00 |
186 | DENSE | 3072 | 768 | 4.00 | 5.24 | 0.06 | 12.78 | 2.44 | 0.17 | 0 | 24.62 | 0.00 |
192 | DENSE | 768 | 768 | 1.00 | 3.37 | 0.09 | 7.50 | 2.23 | 0.18 | 0 | 17.29 | 1.00 |
193 | DENSE | 768 | 768 | 1.00 | 13.18 | 0.10 | 20.77 | 1.58 | 0.16 | 0 | 77.81 | 3.00 |
194 | DENSE | 768 | 768 | 1.00 | 6.86 | 0.05 | 7.94 | 1.16 | 0.04 | 0 | 67.11 | 3.00 |
197 | DENSE | 768 | 768 | 1.00 | 5.25 | 0.07 | 9.13 | 1.74 | 0.14 | 0 | 13.30 | 1.00 |
201 | DENSE | 3072 | 768 | 4.00 | 3.50 | 0.02 | 9.33 | 2.66 | 0.17 | 1 | 11.43 | 0.00 |
204 | DENSE | 3072 | 768 | 4.00 | 4.69 | 0.08 | 12.43 | 2.65 | 0.21 | 0 | 11.68 | 0.00 |
210 | DENSE | 768 | 768 | 1.00 | 2.44 | 0.08 | 4.97 | 2.03 | 0.17 | 0 | 28.89 | 2.00 |
211 | DENSE | 768 | 768 | 1.00 | 5.95 | 0.10 | 10.22 | 1.72 | 0.12 | 0 | 66.87 | 2.00 |
212 | DENSE | 768 | 768 | 1.00 | 10.37 | 0.15 | 9.25 | 0.89 | 0.03 | 0 | 168.65 | 3.00 |
215 | DENSE | 768 | 768 | 1.00 | 4.26 | 0.08 | 7.92 | 1.86 | 0.16 | 1 | 11.22 | 1.00 |
219 | DENSE | 3072 | 768 | 4.00 | 2.67 | 0.03 | 6.32 | 2.37 | 0.21 | 1 | 21.68 | 0.00 |
222 | DENSE | 3072 | 768 | 4.00 | 3.71 | 0.09 | 10.35 | 2.79 | 0.30 | 7 | 5.40 | 0.00 |
226 | DENSE | 768 | 768 | 1.00 | 11.74 | 0.18 | 0.99 | 0.08 | 0.00 | 0 | 194.76 | 2.00 |