File size: 856 Bytes
28afba2 b0abc3f 28afba2 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 |
---
library_name: peft
tags:
- llama1-7b
- code
- instruct
- alpaca-instruct
- alpaca
- llama7b
datasets:
- tatsu-lab/alpaca
base_model: decapoda-research/llama-7b-hf
---
We finetuned huggyllama/llama-7b on tatsu-lab/alpaca Dataset for 5 epochs or ~ 25,000 steps using [MonsterAPI](https://monsterapi.ai) no-code [LLM finetuner](https://docs.monsterapi.ai/fine-tune-a-large-language-model-llm).
This dataset is HuggingFaceH4/tatsu-lab/alpaca unfiltered, removing 36 instances of blatant alignment.
The finetuning session got completed in 4 hours and costed us only `$16` for the entire finetuning run!
#### Hyperparameters & Run details:
- Model Path: huggyllama/llama-7b
- Dataset: tatsu-lab/alpaca
- Learning rate: 0.0003
- Number of epochs: 5
- Data split: Training: 90% / Validation: 10%
- Gradient accumulation steps: 1
license: apache-2.0
---
|