Pruning on Pythia 1B (16L), 410M (24L) (val)
4.4ΔPPLLast-N
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Last-NModel=410M (24L), k=12026.04 | 4.4 | — | |
| WorstModel=410M (24L), k=12026.04 | 4.4 | — | |
| RandomModel=410M (24L), k=12026.04 | 4.43 | — | |
| RandomModel=1B (16L), k=12026.04 | 4.87 | — | |
| Zone-AwareModel=1B (16L), k=12026.04 | 5.6 | 1.1 | |
| Last-NModel=1B (16L), k=12026.04 | 6.1 | — | |
| WorstModel=410M (24L), k=22026.04 | 7 | — | |
| Zone-AwareModel=410M (24L), k=12026.04 | 7.34 | 0.6 | |
| Last-NModel=410M (24L), k=22026.04 | 10.12 | — | |
| RandomModel=410M (24L), k=22026.04 | 11.11 | — | |
| RandomModel=1B (16L), k=22026.04 | 11.61 | — | |
| Zone-AwareModel=1B (16L), k=22026.04 | 11.83 | 1.3 | |
| Last-NModel=1B (16L), k=22026.04 | 15.23 | — | |
| Zone-AwareModel=1B (16L), k=32026.04 | 20.9 | 2.6 | |
| Zone-AwareModel=410M (24L), k=22026.04 | 23.77 | 0.4 | |
| RandomModel=1B (16L), k=32026.04 | 27.95 | — | |
| WorstModel=410M (24L), k=42026.04 | 33.08 | — | |
| WorstModel=1B (16L), k=12026.04 | 40.4 | — | |
| Last-NModel=410M (24L), k=42026.04 | 41.89 | — | |
| Last-NModel=1B (16L), k=32026.04 | 53.36 | — | |
| Zone-AwareModel=1B (16L), k=42026.04 | 55.32 | 3.6 | |
| Zone-AwareModel=410M (24L), k=42026.04 | 83.2 | 0.5 | |
| WorstModel=1B (16L), k=22026.04 | 86.58 | — | |
| WorstModel=410M (24L), k=62026.04 | 99.23 | — | |
| RandomModel=1B (16L), k=42026.04 | 117.92 | — | |
| RandomModel=410M (24L), k=42026.04 | 121.21 | — | |
| Last-NModel=1B (16L), k=42026.04 | 199.12 | — | |
| Zone-AwareModel=410M (24L), k=62026.04 | 258.24 | 1.2 | |
| Last-NModel=410M (24L), k=62026.04 | 310.01 | — | |
| RandomModel=410M (24L), k=62026.04 | 320.82 | — | |
| Zone-AwareModel=410M (24L), k=82026.04 | 707.73 | 2 | |
| Last-NModel=410M (24L), k=82026.04 | 1,383.35 | — | |
| RandomModel=410M (24L), k=82026.04 | 2,129 | — | |
| WorstModel=1B (16L), k=32026.04 | 11,703 | — | |
| WorstModel=1B (16L), k=42026.04 | 15,146 | — | |
| WorstModel=410M (24L), k=82026.04 | 29,548 | — |