Metadata: | |
AR loss to consistency loss ratio: 10: 1 | |
shareGPT dataset size: 48k | |
n-token sequence length: 32 | |
Jacobi trajectory data cleaning: True | |
Target model: LLaMA2-7B fine-tuned on ShareGPT48k | |
release date: 02/26/2024 |
Metadata: | |
AR loss to consistency loss ratio: 10: 1 | |
shareGPT dataset size: 48k | |
n-token sequence length: 32 | |
Jacobi trajectory data cleaning: True | |
Target model: LLaMA2-7B fine-tuned on ShareGPT48k | |
release date: 02/26/2024 |