Skip to content

Update examples training script path links #52

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 2 commits into
base: main
Choose a base branch
from
Open

Update examples training script path links #52

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
87a055e
update script path links
guspan-tanadi Apr 18, 2025
1cf34f9
update Embeddings path link
guspan-tanadi Apr 18, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
12 changes: 6 additions & 6 deletions docs/finetune-embedding.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -23,8 +23,8 @@ Here are some common tasks and their associated training methods

| | Dataset Format | Loss | Evaluator |
|-------------------------------|----------------|---------------|---------------|
|**Duplicate Text Classification [[source]](https://github.com/UKPLab/sentence-transformers/blob/master/examples/training/quora_duplicate_questions/training_OnlineContrastiveLoss.py)**|Pairs of text that are labeled as duplicate or not|[Online Contrastive Loss](https://www.sbert.net/docs/package_reference/losses.html#onlinecontrastiveloss)|[Binary Classification Evaluation](https://www.sbert.net/docs/package_reference/evaluation.html#sentence_transformers.evaluation.BinaryClassificationEvaluator)|
|**Information Retrieval [[source]](https://github.com/UKPLab/sentence-transformers/blob/master/examples/training/quora_duplicate_questions/training_MultipleNegativesRankingLoss.py)**|Pairs of text that are semantically relevant (don't need labeled duplicates)|[Multiple Negatives Ranking Loss](https://www.sbert.net/docs/package_reference/losses.html#multiplenegativesrankingloss)|[Information Retrieval Evaluator](https://www.sbert.net/docs/package_reference/evaluation.html#sentence_transformers.evaluation.InformationRetrievalEvaluator)|
|**Duplicate Text Classification [[source]](https://github.com/UKPLab/sentence-transformers/blob/master/examples/sentence_transformer/training/quora_duplicate_questions/training_OnlineContrastiveLoss.py)**|Pairs of text that are labeled as duplicate or not|[Online Contrastive Loss](https://www.sbert.net/docs/package_reference/losses.html#onlinecontrastiveloss)|[Binary Classification Evaluation](https://www.sbert.net/docs/package_reference/evaluation.html#sentence_transformers.evaluation.BinaryClassificationEvaluator)|
|**Information Retrieval [[source]](https://github.com/UKPLab/sentence-transformers/blob/master/examples/sentence_transformer/training/quora_duplicate_questions/training_MultipleNegativesRankingLoss.py)**|Pairs of text that are semantically relevant (don't need labeled duplicates)|[Multiple Negatives Ranking Loss](https://www.sbert.net/docs/package_reference/losses.html#multiplenegativesrankingloss)|[Information Retrieval Evaluator](https://www.sbert.net/docs/package_reference/evaluation.html#sentence_transformers.evaluation.InformationRetrievalEvaluator)|

For this article, we will be choosing the specific task of **Duplicate Text Classification**, i.e. given two questions, identify whether they are semantically duplicates. Let's go over which **dataset format**, **loss**, and **evaluator** to use for this specific fine-tuning task.

Expand Down Expand Up @@ -115,7 +115,7 @@ Nice! 👍 Our dataset is now in the correct format for training.

The loss function we choose is also dependent on our specific task. In our case, we will be using `OnlineContrastiveLoss` since it fits our task of duplicate pair classification.

A detailed explanation can be [found here](https://www.sbert.net/examples/training/quora_duplicate_questions/README.html#constrative-loss), but in short contrastive loss functions work by optimizing for the condition where "Similar pairs with label 1 are pulled together, so that they are close in vector space. Dissimilar pairs, that are closer than a defined margin, are pushed away in vector space." (SBERT)
A detailed explanation can be [found here](https://www.sbert.net/examples/training/quora_duplicate_questions/README.html#contrastive-loss), but in short contrastive loss functions work by optimizing for the condition where "Similar pairs with label 1 are pulled together, so that they are close in vector space. Dissimilar pairs, that are closer than a defined margin, are pushed away in vector space." (SBERT)

We initialize our loss function as shown

Expand All @@ -129,7 +129,7 @@ train_loss = losses.OnlineContrastiveLoss(model)

The evaluation function we choose is also dependent on our specific task. In our case of duplicate pair classification, we use `BinaryClassificationEvaluator`.

`BinaryClassificationEvaluator` works by the question evaluating "Given (question1, question2), is this a duplicate or not?"[[SBERT code]](https://github.com/UKPLab/sentence-transformers/blob/master/examples/training/quora_duplicate_questions/training_OnlineContrastiveLoss.py#L78-L93).
`BinaryClassificationEvaluator` works by the question evaluating "Given (question1, question2), is this a duplicate or not?"[[SBERT code]](https://github.com/UKPLab/sentence-transformers/blob/master/examples/sentence_transformer/training/quora_duplicate_questions/training_OnlineContrastiveLoss.py#L60-L71).

We initialize the evaluator function using our test data

Expand Down Expand Up @@ -215,8 +215,8 @@ Here are some common tasks and their associated training methods

| | Dataset Format | Loss | Evaluator |
|-------------------------------|----------------|---------------|---------------|
|**Duplicate Text Classification [src](https://github.com/UKPLab/sentence-transformers/blob/master/examples/training/quora_duplicate_questions/training_OnlineContrastiveLoss.py)**|Pairs of text that are may or may not be duplicate|[OnlineContrastiveLoss](https://www.sbert.net/docs/package_reference/losses.html#onlinecontrastiveloss)|[BinaryClassificationEvaluation](https://www.sbert.net/docs/package_reference/evaluation.html#sentence_transformers.evaluation.BinaryClassificationEvaluator)|
|**Information Retrieval [src](https://github.com/UKPLab/sentence-transformers/blob/master/examples/training/quora_duplicate_questions/training_MultipleNegativesRankingLoss.py)**|Pairs of text that are semantically relevant (don't need negatives)|[MultipleNegativesRankingLoss](https://www.sbert.net/docs/package_reference/losses.html#multiplenegativesrankingloss)|[InformationRetrievalEvaluator](https://www.sbert.net/docs/package_reference/evaluation.html#sentence_transformers.evaluation.InformationRetrievalEvaluator)|
|**Duplicate Text Classification [src](https://github.com/UKPLab/sentence-transformers/blob/master/examples/sentence_transformer/training/quora_duplicate_questions/training_OnlineContrastiveLoss.py)**|Pairs of text that are may or may not be duplicate|[OnlineContrastiveLoss](https://www.sbert.net/docs/package_reference/losses.html#onlinecontrastiveloss)|[BinaryClassificationEvaluation](https://www.sbert.net/docs/package_reference/evaluation.html#sentence_transformers.evaluation.BinaryClassificationEvaluator)|
|**Information Retrieval [src](https://github.com/UKPLab/sentence-transformers/blob/master/examples/sentence_transformer/training/quora_duplicate_questions/training_MultipleNegativesRankingLoss.py)**|Pairs of text that are semantically relevant (don't need negatives)|[MultipleNegativesRankingLoss](https://www.sbert.net/docs/package_reference/losses.html#multiplenegativesrankingloss)|[InformationRetrievalEvaluator](https://www.sbert.net/docs/package_reference/evaluation.html#sentence_transformers.evaluation.InformationRetrievalEvaluator)|

## Fine-Tuning on Modal

Expand Down
4 changes: 2 additions & 2 deletions docs/index.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -8,7 +8,7 @@ The goal of this repository is to provide a series of Docker containers, or moda

## Roadmap

- [x] Support [GPT4All's Embedding API](https://docs.gpt4all.io/gpt4all_python_embedding.html) and match it to [OpenAI Embeddings](https://openai.com/v1/embedding)
- [x] Support [GPT4All's Embedding API](https://docs.gpt4all.io/gpt4all_python_embedding.html) and match it to [OpenAI Embeddings](https://platform.openai.com/docs/api-reference/embeddings)
- [ ] Support JSONFormer API to match it to `ChatCompletion` with `function_calls`
- [ ] Support Cross Encoders based on sentence transformers for any Hugging Face model
- [ ] Provide great documentation and runbooks using MkDocs
Expand All @@ -19,4 +19,4 @@ Contributions are welcome! If you have any suggestions, improvements, or bug fix

## License

This project is licensed under the [MIT License](https://opensource.org/licenses/MIT).
This project is licensed under the [MIT License](https://opensource.org/licenses/MIT).