https://github.com/eliask93/instruction-fine-tuned-gemma-2-for-stance-detection

Example application for applying QLoRA-based Parameter-Efficient Fine-Tuning (PEFT) to a Stance Detection task using Gemma-2-9B-Instruct
https://github.com/eliask93/instruction-fine-tuned-gemma-2-for-stance-detection

argument-mining gemma2 lora nlp peft qlora quantization stance-detection

Last synced: 8 months ago
JSON representation

Example application for applying QLoRA-based Parameter-Efficient Fine-Tuning (PEFT) to a Stance Detection task using Gemma-2-9B-Instruct

• Host: GitHub
• URL: https://github.com/eliask93/instruction-fine-tuned-gemma-2-for-stance-detection
• Owner: EliasK93
• Created: 2024-09-06T18:52:14.000Z (about 1 year ago)
• Default Branch: master
• Last Pushed: 2024-09-06T19:20:03.000Z (about 1 year ago)
• Last Synced: 2025-01-12T15:46:09.145Z (9 months ago)
• Topics: argument-mining, gemma2, lora, nlp, peft, qlora, quantization, stance-detection
• Language: Python
• Homepage:
• Size: 1.98 MB
• Stars: 1
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

          
## Parameter-Efficient Fine-Tuning of Gemma-2-Instruct-9B for Stance Detection 

Example application for fine-tuning the instruction-tuned 9B variant of Google's [Gemma 2](https://blog.google/technology/developers/google-gemma-2/) model ([google/gemma-2-9b-it](https://huggingface.co/google/gemma-2-9b-it)) on a new task (Stance Detection). 

To make the fine-tuning feasible on a consumer GPU, a parameter-efficient fine-tuning (PEFT) approach based on [QLoRA](https://arxiv.org/abs/2305.14314) (_Quantized Low-Rank Adaptation_) is applied. While the original model has 9.295.724.032 parameters, this approach works with only 54.018.048 trainable parameters (~0.581%), leaving the original model weights entirely frozen during the fine-tuning process. This way, the model was fine-tuned for two epochs on the task using a single RTX 4080 GPU.




### Corpus

Each model was fine-tuned on a 5,000 sentences Stance Detection corpus that I manually annotated during my Master's Thesis.

Stance Detection aims to classify the stance a sentence takes towards a claim (topic) as either _Pro_, _Contra_ or _Neutral_.

The sentences originate from Reddit's _r/ChangeMyView_ subreddit in the time span between January 2013 and October 2018, as provided in the [ConvoKit subreddit corpus](https://zissou.infosci.cornell.edu/convokit/datasets/subreddit-corpus/corpus-zipped/).

They cover five topics: _abortion_, _climate change_, _gun control_, _minimum wage_ and _veganism_.

The table below shows some examples.

topic

sentence

stance label

There should be more gun control.

It's the only country with a "2nd Amendment", yet 132 countries have a lower murder rate.

Pro

Humanity needs to combat climate change.

The overhwelming evidence could be lies and you would never know because you're content to live your life as a giant appeal to authority.

Contra

Vegans are right.

It's all about finding a system that works for you.

Neutral




For the instruction-based fine-tuning and inference, the sentence pairs are wrapped in the following prompt:

prompt

expected output

The first text is a hypothesis/claim, the second text is a sentence. Determine whether the sentence is a pro argument ("pro"), a contra argument ("con") or doesnt take position at all/is neutral ("neu") towards the hypothesis.


For your answer, just write exactly one of pro, con or neu, not a full text.


Sentence to predict:


hypothesis: \`\`\`There should be more gun control.\`\`\`


sentence: \`\`\`It's the only country with a "2nd Amendment", yet 132 countries have a lower murder rate.\`\`\`


stance:

pro




### Results

|                      Model                       | Accuracy | Micro-F1 | Macro-F1 |

|:------------------------------------------------:|:--------:|:--------:|:--------:|

|        base model (google/gemma-2-9b-it)         |   0.72   |   0.71   |   0.71   |

| fine-tuned model (gemma2-9b-it-stance-finetuned) |   0.90   |   0.89   |   0.90   |




### Requirements

##### - Python >= 3.10

##### - Conda

  - `pytorch==2.4.0`

  - `cudatoolkit=12.1`

##### - pip

  - `transformers`

  - `datasets`

  - `trl`

  - `sentencepiece`

  - `protobuf`

  - `peft`

  - `bitsandbytes`

  - `openpyxl`

  - `scikit-learn`




### Notes

The dataset files in this repository are cut off after the first 50 rows.

The trained model files `adapter_model.safetensors`, `optimizer.pt` and `tokenizer.json` are omitted in this repository.