File size: 997 Bytes
dba9c06 a7580a4 dba9c06 a7580a4 |
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 36 37 |
---
datasets:
- ewof/code-alpaca-instruct-unfiltered
library_name: peft
tags:
- gpt-j
- gpt-j-6b
- code
- instruct
- instruct-code
- code-alpaca
- alpaca-instruct
- alpaca
- llama7b
- gpt2
---
We finetuned GPT-J 6B on Code-Alpaca-Instruct Dataset (ewof/code-alpaca-instruct-unfiltered) 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 206 minutes and costed us only `$8` for the entire finetuning run!
#### Hyperparameters & Run details:
- Model Path: EleutherAI/gpt-j-6b
- Dataset: ewof/code-alpaca-instruct-unfiltered
- Learning rate: 0.0003
- Number of epochs: 5
- Data split: Training: 90% / Validation: 10%
- Gradient accumulation steps: 1
Loss metrics:
---
license: apache-2.0
--- |