layer_type | N | M | Q | alpha | D | alpha-hat | log_SN | rand_D | num_traps | stable_rank | rank_loss | |
---|---|---|---|---|---|---|---|---|---|---|---|---|
layer_id | ||||||||||||
2 | EMBEDDING | 50265 | 768 | 65.45 | 2.91 | 0.05 | 12.66 | 4.35 | 0.24 | 1 | 30.34 | 0.00 |
3 | EMBEDDING | 768 | 514 | 1.49 | 1.51 | 0.10 | 4.31 | 2.86 | 0.72 | 0 | 2.38 | 1.00 |
12 | DENSE | 768 | 768 | 1.00 | 3.54 | 0.04 | 7.15 | 2.02 | 0.12 | 0 | 58.82 | 0.00 |
13 | DENSE | 768 | 768 | 1.00 | 2.89 | 0.07 | 6.10 | 2.11 | 0.18 | 0 | 46.84 | 2.00 |
14 | DENSE | 768 | 768 | 1.00 | 3.37 | 0.12 | 3.40 | 1.01 | 0.12 | 0 | 83.50 | 1.00 |
17 | DENSE | 768 | 768 | 1.00 | 4.20 | 0.05 | 6.93 | 1.65 | 0.18 | 0 | 19.30 | 1.00 |
21 | DENSE | 3072 | 768 | 4.00 | 2.77 | 0.03 | 7.36 | 2.66 | 0.19 | 1 | 16.16 | 0.00 |
24 | DENSE | 3072 | 768 | 4.00 | 3.48 | 0.03 | 6.71 | 1.93 | 0.14 | 0 | 56.94 | 0.00 |
30 | DENSE | 768 | 768 | 1.00 | 2.86 | 0.05 | 6.27 | 2.19 | 0.16 | 0 | 21.32 | 1.00 |
31 | DENSE | 768 | 768 | 1.00 | 2.68 | 0.05 | 4.99 | 1.86 | 0.18 | 0 | 43.59 | 1.00 |
32 | DENSE | 768 | 768 | 1.00 | 5.65 | 0.10 | 6.86 | 1.21 | 0.15 | 0 | 63.08 | 1.00 |
35 | DENSE | 768 | 768 | 1.00 | 2.49 | 0.08 | 3.26 | 1.31 | 0.17 | 0 | 41.78 | 1.00 |
39 | DENSE | 3072 | 768 | 4.00 | 3.15 | 0.02 | 8.79 | 2.79 | 0.15 | 1 | 13.97 | 0.00 |
42 | DENSE | 3072 | 768 | 4.00 | 3.31 | 0.05 | 6.14 | 1.85 | 0.15 | 0 | 72.90 | 0.00 |
48 | DENSE | 768 | 768 | 1.00 | 3.26 | 0.06 | 7.01 | 2.15 | 0.16 | 0 | 19.01 | 1.00 |
49 | DENSE | 768 | 768 | 1.00 | 2.97 | 0.04 | 5.05 | 1.70 | 0.16 | 0 | 52.75 | 1.00 |
50 | DENSE | 768 | 768 | 1.00 | 5.32 | 0.07 | 7.40 | 1.39 | 0.14 | 0 | 51.02 | 2.00 |
53 | DENSE | 768 | 768 | 1.00 | 3.59 | 0.09 | 4.36 | 1.21 | 0.13 | 0 | 57.65 | 1.00 |
57 | DENSE | 3072 | 768 | 4.00 | 3.82 | 0.02 | 10.51 | 2.75 | 0.13 | 1 | 15.92 | 0.00 |
60 | DENSE | 3072 | 768 | 4.00 | 4.54 | 0.05 | 8.67 | 1.91 | 0.15 | 0 | 67.96 | 0.00 |
66 | DENSE | 768 | 768 | 1.00 | 3.48 | 0.06 | 7.74 | 2.23 | 0.16 | 0 | 17.97 | 2.00 |
67 | DENSE | 768 | 768 | 1.00 | 3.46 | 0.06 | 5.74 | 1.66 | 0.12 | 0 | 62.06 | 1.00 |
68 | DENSE | 768 | 768 | 1.00 | 5.13 | 0.05 | 8.14 | 1.59 | 0.14 | 0 | 33.79 | 3.00 |
71 | DENSE | 768 | 768 | 1.00 | 3.37 | 0.07 | 4.12 | 1.22 | 0.13 | 0 | 65.07 | 1.00 |
75 | DENSE | 3072 | 768 | 4.00 | 3.86 | 0.03 | 10.39 | 2.69 | 0.14 | 1 | 17.74 | 0.00 |
78 | DENSE | 3072 | 768 | 4.00 | 4.28 | 0.04 | 8.86 | 2.07 | 0.15 | 0 | 44.78 | 0.00 |
84 | DENSE | 768 | 768 | 1.00 | 3.45 | 0.04 | 7.67 | 2.22 | 0.16 | 0 | 18.95 | 2.00 |
85 | DENSE | 768 | 768 | 1.00 | 3.37 | 0.06 | 5.66 | 1.68 | 0.13 | 0 | 61.14 | 1.00 |
86 | DENSE | 768 | 768 | 1.00 | 5.46 | 0.07 | 8.21 | 1.50 | 0.16 | 0 | 42.82 | 1.00 |
89 | DENSE | 768 | 768 | 1.00 | 3.78 | 0.08 | 4.61 | 1.22 | 0.13 | 0 | 63.95 | 2.00 |
93 | DENSE | 3072 | 768 | 4.00 | 3.84 | 0.03 | 10.15 | 2.64 | 0.14 | 1 | 17.38 | 0.00 |
96 | DENSE | 3072 | 768 | 4.00 | 4.69 | 0.05 | 9.10 | 1.94 | 0.15 | 0 | 52.14 | 0.00 |
102 | DENSE | 768 | 768 | 1.00 | 3.90 | 0.06 | 8.77 | 2.25 | 0.16 | 0 | 15.87 | 1.00 |
103 | DENSE | 768 | 768 | 1.00 | 4.16 | 0.08 | 6.60 | 1.59 | 0.11 | 0 | 67.73 | 3.00 |
104 | DENSE | 768 | 768 | 1.00 | 5.34 | 0.11 | 6.90 | 1.29 | 0.14 | 0 | 79.64 | 1.00 |
107 | DENSE | 768 | 768 | 1.00 | 4.04 | 0.07 | 5.44 | 1.35 | 0.15 | 0 | 52.37 | 1.00 |
111 | DENSE | 3072 | 768 | 4.00 | 3.89 | 0.02 | 9.71 | 2.49 | 0.14 | 1 | 21.80 | 0.00 |
114 | DENSE | 3072 | 768 | 4.00 | 4.03 | 0.03 | 7.89 | 1.96 | 0.15 | 0 | 44.20 | 0.00 |
120 | DENSE | 768 | 768 | 1.00 | 6.59 | 0.09 | 10.80 | 1.64 | 0.14 | 0 | 64.38 | 2.00 |
121 | DENSE | 768 | 768 | 1.00 | 2.52 | 0.11 | 3.93 | 1.56 | 0.15 | 0 | 71.40 | 2.00 |
122 | DENSE | 768 | 768 | 1.00 | 8.27 | 0.06 | 10.01 | 1.21 | 0.11 | 0 | 92.01 | 2.00 |
125 | DENSE | 768 | 768 | 1.00 | 4.21 | 0.04 | 6.80 | 1.61 | 0.12 | 0 | 27.92 | 1.00 |
129 | DENSE | 3072 | 768 | 4.00 | 4.03 | 0.03 | 10.66 | 2.65 | 0.13 | 1 | 16.86 | 0.00 |
132 | DENSE | 3072 | 768 | 4.00 | 4.14 | 0.02 | 8.54 | 2.07 | 0.16 | 0 | 38.34 | 0.00 |
138 | DENSE | 768 | 768 | 1.00 | 4.00 | 0.08 | 6.70 | 1.67 | 0.13 | 0 | 68.60 | 2.00 |
139 | DENSE | 768 | 768 | 1.00 | 3.68 | 0.06 | 6.34 | 1.72 | 0.15 | 0 | 57.70 | 2.00 |
140 | DENSE | 768 | 768 | 1.00 | 6.97 | 0.07 | 8.26 | 1.19 | 0.12 | 0 | 70.73 | 1.00 |
143 | DENSE | 768 | 768 | 1.00 | 4.96 | 0.05 | 6.31 | 1.27 | 0.10 | 0 | 42.32 | 1.00 |
147 | DENSE | 3072 | 768 | 4.00 | 3.91 | 0.03 | 10.09 | 2.58 | 0.13 | 1 | 19.84 | 0.00 |
150 | DENSE | 3072 | 768 | 4.00 | 3.81 | 0.04 | 8.42 | 2.21 | 0.17 | 0 | 28.86 | 0.00 |
156 | DENSE | 768 | 768 | 1.00 | 3.93 | 0.06 | 7.22 | 1.84 | 0.15 | 0 | 39.51 | 2.00 |
157 | DENSE | 768 | 768 | 1.00 | 3.62 | 0.08 | 6.04 | 1.67 | 0.16 | 0 | 55.42 | 2.00 |
158 | DENSE | 768 | 768 | 1.00 | 6.14 | 0.04 | 8.19 | 1.34 | 0.06 | 0 | 53.64 | 2.00 |
161 | DENSE | 768 | 768 | 1.00 | 5.68 | 0.06 | 7.49 | 1.32 | 0.07 | 0 | 38.14 | 1.00 |
165 | DENSE | 3072 | 768 | 4.00 | 3.67 | 0.03 | 9.92 | 2.70 | 0.14 | 1 | 14.26 | 0.00 |
168 | DENSE | 3072 | 768 | 4.00 | 3.96 | 0.03 | 8.52 | 2.15 | 0.16 | 0 | 33.47 | 0.00 |
174 | DENSE | 768 | 768 | 1.00 | 4.04 | 0.07 | 7.96 | 1.97 | 0.18 | 0 | 27.69 | 2.00 |
175 | DENSE | 768 | 768 | 1.00 | 3.29 | 0.10 | 5.44 | 1.65 | 0.16 | 0 | 55.41 | 2.00 |
176 | DENSE | 768 | 768 | 1.00 | 4.67 | 0.05 | 7.01 | 1.50 | 0.08 | 0 | 32.07 | 3.00 |
179 | DENSE | 768 | 768 | 1.00 | 5.10 | 0.07 | 6.76 | 1.33 | 0.08 | 0 | 34.44 | 1.00 |
183 | DENSE | 3072 | 768 | 4.00 | 3.64 | 0.03 | 9.65 | 2.65 | 0.14 | 1 | 12.90 | 0.00 |
186 | DENSE | 3072 | 768 | 4.00 | 4.68 | 0.04 | 9.72 | 2.08 | 0.12 | 0 | 37.68 | 0.00 |
192 | DENSE | 768 | 768 | 1.00 | 4.75 | 0.08 | 10.09 | 2.12 | 0.17 | 0 | 19.82 | 1.00 |
193 | DENSE | 768 | 768 | 1.00 | 6.05 | 0.14 | 8.84 | 1.46 | 0.15 | 0 | 84.78 | 2.00 |
194 | DENSE | 768 | 768 | 1.00 | 5.56 | 0.06 | 7.87 | 1.41 | 0.07 | 0 | 39.00 | 3.00 |
197 | DENSE | 768 | 768 | 1.00 | 4.97 | 0.07 | 6.84 | 1.38 | 0.08 | 0 | 31.87 | 1.00 |
201 | DENSE | 3072 | 768 | 4.00 | 3.69 | 0.03 | 9.09 | 2.47 | 0.14 | 1 | 15.74 | 0.00 |
204 | DENSE | 3072 | 768 | 4.00 | 4.85 | 0.08 | 10.39 | 2.14 | 0.12 | 0 | 28.68 | 0.00 |
210 | DENSE | 768 | 768 | 1.00 | 4.54 | 0.04 | 9.54 | 2.10 | 0.17 | 0 | 21.82 | 2.00 |
211 | DENSE | 768 | 768 | 1.00 | 7.23 | 0.05 | 11.39 | 1.58 | 0.12 | 0 | 70.65 | 2.00 |
212 | DENSE | 768 | 768 | 1.00 | 7.61 | 0.10 | 9.82 | 1.29 | 0.05 | 0 | 66.03 | 3.00 |
215 | DENSE | 768 | 768 | 1.00 | 5.13 | 0.08 | 7.02 | 1.37 | 0.07 | 0 | 37.41 | 0.00 |
219 | DENSE | 3072 | 768 | 4.00 | 3.57 | 0.02 | 9.15 | 2.56 | 0.13 | 1 | 13.65 | 0.00 |
222 | DENSE | 3072 | 768 | 4.00 | 4.66 | 0.09 | 10.84 | 2.33 | 0.18 | 0 | 13.62 | 0.00 |
226 | DENSE | 768 | 768 | 1.00 | 45.11 | 0.17 | 3.98 | 0.09 | 0.00 | 0 | 192.40 | 1.00 |