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 | 3.23 | 0.05 | 14.17 | 4.39 | 0.24 | 1 | 26.63 | 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 | 3.31 | 0.04 | 6.99 | 2.11 | 0.13 | 0 | 49.61 | 1.00 |
13 | DENSE | 768 | 768 | 1.00 | 2.81 | 0.07 | 6.06 | 2.16 | 0.20 | 0 | 43.82 | 1.00 |
14 | DENSE | 768 | 768 | 1.00 | 2.71 | 0.12 | 2.95 | 1.09 | 0.13 | 0 | 76.31 | 1.00 |
17 | DENSE | 768 | 768 | 1.00 | 4.13 | 0.05 | 6.80 | 1.65 | 0.17 | 0 | 21.35 | 1.00 |
21 | DENSE | 3072 | 768 | 4.00 | 2.80 | 0.02 | 7.46 | 2.66 | 0.19 | 1 | 16.80 | 0.00 |
24 | DENSE | 3072 | 768 | 4.00 | 3.31 | 0.04 | 6.56 | 1.98 | 0.15 | 0 | 53.94 | 0.00 |
30 | DENSE | 768 | 768 | 1.00 | 3.76 | 0.05 | 8.30 | 2.21 | 0.17 | 0 | 18.81 | 2.00 |
31 | DENSE | 768 | 768 | 1.00 | 2.87 | 0.06 | 5.11 | 1.78 | 0.17 | 0 | 48.85 | 1.00 |
32 | DENSE | 768 | 768 | 1.00 | 7.00 | 0.08 | 9.86 | 1.41 | 0.15 | 0 | 53.78 | 0.00 |
35 | DENSE | 768 | 768 | 1.00 | 4.98 | 0.09 | 6.60 | 1.32 | 0.13 | 0 | 50.69 | 2.00 |
39 | DENSE | 3072 | 768 | 4.00 | 3.58 | 0.02 | 10.05 | 2.81 | 0.15 | 1 | 14.68 | 0.00 |
42 | DENSE | 3072 | 768 | 4.00 | 5.14 | 0.05 | 10.36 | 2.02 | 0.17 | 0 | 56.41 | 0.00 |
48 | DENSE | 768 | 768 | 1.00 | 3.92 | 0.04 | 8.85 | 2.25 | 0.17 | 0 | 18.28 | 1.00 |
49 | DENSE | 768 | 768 | 1.00 | 4.22 | 0.09 | 7.12 | 1.69 | 0.14 | 0 | 62.65 | 2.00 |
50 | DENSE | 768 | 768 | 1.00 | 5.85 | 0.06 | 9.25 | 1.58 | 0.16 | 0 | 41.63 | 1.00 |
53 | DENSE | 768 | 768 | 1.00 | 3.59 | 0.10 | 4.64 | 1.29 | 0.13 | 0 | 64.32 | 1.00 |
57 | DENSE | 3072 | 768 | 4.00 | 3.69 | 0.02 | 9.97 | 2.70 | 0.16 | 1 | 16.31 | 0.00 |
60 | DENSE | 3072 | 768 | 4.00 | 4.67 | 0.04 | 9.49 | 2.03 | 0.17 | 0 | 45.09 | 0.00 |
66 | DENSE | 768 | 768 | 1.00 | 4.04 | 0.09 | 7.06 | 1.75 | 0.14 | 0 | 60.62 | 1.00 |
67 | DENSE | 768 | 768 | 1.00 | 3.66 | 0.07 | 6.36 | 1.74 | 0.17 | 0 | 58.23 | 2.00 |
68 | DENSE | 768 | 768 | 1.00 | 8.71 | 0.08 | 11.27 | 1.29 | 0.13 | 0 | 69.27 | 1.00 |
71 | DENSE | 768 | 768 | 1.00 | 5.00 | 0.06 | 6.56 | 1.31 | 0.10 | 0 | 50.53 | 1.00 |
75 | DENSE | 3072 | 768 | 4.00 | 3.56 | 0.03 | 9.53 | 2.68 | 0.16 | 1 | 16.42 | 0.00 |
78 | DENSE | 3072 | 768 | 4.00 | 3.59 | 0.02 | 8.00 | 2.23 | 0.19 | 0 | 29.84 | 0.00 |
84 | DENSE | 768 | 768 | 1.00 | 3.54 | 0.08 | 7.26 | 2.05 | 0.20 | 0 | 26.40 | 2.00 |
85 | DENSE | 768 | 768 | 1.00 | 4.32 | 0.11 | 7.26 | 1.68 | 0.19 | 0 | 58.21 | 2.00 |
86 | DENSE | 768 | 768 | 1.00 | 5.18 | 0.06 | 8.21 | 1.59 | 0.08 | 0 | 35.18 | 2.00 |
89 | DENSE | 768 | 768 | 1.00 | 3.48 | 0.06 | 4.87 | 1.40 | 0.07 | 0 | 41.38 | 1.00 |
93 | DENSE | 3072 | 768 | 4.00 | 3.39 | 0.02 | 9.21 | 2.72 | 0.16 | 1 | 11.23 | 0.00 |
96 | DENSE | 3072 | 768 | 4.00 | 4.43 | 0.04 | 9.36 | 2.12 | 0.13 | 0 | 36.19 | 0.00 |
102 | DENSE | 768 | 768 | 1.00 | 4.33 | 0.07 | 9.31 | 2.15 | 0.20 | 0 | 19.56 | 1.00 |
103 | DENSE | 768 | 768 | 1.00 | 8.99 | 0.08 | 14.38 | 1.60 | 0.17 | 0 | 66.55 | 2.00 |
104 | DENSE | 768 | 768 | 1.00 | 7.08 | 0.09 | 9.95 | 1.41 | 0.05 | 0 | 63.27 | 2.00 |
107 | DENSE | 768 | 768 | 1.00 | 4.90 | 0.10 | 7.08 | 1.44 | 0.07 | 0 | 43.07 | 1.00 |
111 | DENSE | 3072 | 768 | 4.00 | 3.34 | 0.02 | 8.77 | 2.62 | 0.17 | 1 | 11.97 | 0.00 |
114 | DENSE | 3072 | 768 | 4.00 | 4.69 | 0.08 | 10.82 | 2.31 | 0.18 | 0 | 14.94 | 0.00 |
118 | DENSE | 768 | 768 | 1.00 | 9.92 | 0.18 | 0.84 | 0.08 | 0.00 | 0 | 193.83 | 1.00 |