Language Modeling on TinyLlama 1.1B (test)
54.15PerplexityDonor
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| DonorModel=TinyLlama (1.1B), Trainable %=18.0%2026.01 | 54.15 | 8.1 | — | |
| LoRAModel=TinyLlama (1.1B), Trainable %=0.41%2026.01 | 460.41 | 116.5 | 14.4 | |
| Full fine-tuneModel=TinyLlama (1.1B), Trainable %=100%2026.01 | 788.36 | 104.8 | 12.9 | |
| Top-KModel=TinyLlama (1.1B), Trainable %=18.0%2026.01 | 1,157.83 | 41.3 | 5.1 | |
| Zero-shotModel=TinyLlama (1.1B), Trainable %=0%2026.01 | 1,493.16 | — | — | |
| IA3Model=TinyLlama (1.1B), Trainable %=0.01%2026.01 | 2,019.41 | 86 | 10.6 |