File size: 937 Bytes
68d50e2 170e82a 499947b e502cae 68d50e2 170e82a 499947b 170e82a 499947b 170e82a 499947b |
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 32 33 34 35 |
---
library_name: peft
tags:
- codellama7b
- code
- instruct
- instruct-code
- code-alpaca
- alpaca-instruct
- alpaca
- codellama7b
- gpt2
datasets:
- sahil2801/CodeAlpaca-20k
base_model: codellama/CodeLlama-7b-hf
---
We finetuned CodeLlama7B on Code-Alpaca-Instruct Dataset (sahil2801/CodeAlpaca-20k) 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/CodeAlpaca_20K 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: meta-llama/CodeLlama7B
- Dataset: sahil2801/CodeAlpaca-20k
- Learning rate: 0.0003
- Number of epochs: 5
- Data split: Training: 90% / Validation: 10%
- Gradient accumulation steps: 1
---
license: apache-2.0
---
|