|
--- |
|
license: apache-2.0 |
|
pipeline_tag: sentence-similarity |
|
tags: |
|
- sentence-transformers |
|
- feature-extraction |
|
- sentence-similarity |
|
- transformers |
|
- generated_from_trainer |
|
datasets: |
|
- squad |
|
- newsqa |
|
- LLukas22/cqadupstack |
|
- LLukas22/fiqa |
|
- LLukas22/scidocs |
|
- deepset/germanquad |
|
- LLukas22/nq |
|
language: |
|
- en |
|
--- |
|
|
|
# all-mpnet-base-v2-embedding-all |
|
|
|
This model is a fine-tuned version of [all-mpnet-base-v2](https://huggingface.co/sentence-transformers/all-mpnet-base-v2) on the following datasets: [squad](https://huggingface.co/datasets/squad), [newsqa](https://huggingface.co/datasets/newsqa), [LLukas22/cqadupstack](https://huggingface.co/datasets/LLukas22/cqadupstack), [LLukas22/fiqa](https://huggingface.co/datasets/LLukas22/fiqa), [LLukas22/scidocs](https://huggingface.co/datasets/LLukas22/scidocs), [deepset/germanquad](https://huggingface.co/datasets/deepset/germanquad), [LLukas22/nq](https://huggingface.co/datasets/LLukas22/nq). |
|
|
|
|
|
|
|
## Usage (Sentence-Transformers) |
|
|
|
Using this model becomes easy when you have [sentence-transformers](https://www.SBERT.net) installed: |
|
|
|
``` |
|
pip install -U sentence-transformers |
|
``` |
|
|
|
Then you can use the model like this: |
|
|
|
```python |
|
from sentence_transformers import SentenceTransformer |
|
sentences = ["This is an example sentence", "Each sentence is converted"] |
|
|
|
model = SentenceTransformer('LLukas22/all-mpnet-base-v2-embedding-all') |
|
embeddings = model.encode(sentences) |
|
print(embeddings) |
|
``` |
|
|
|
## Training hyperparameters |
|
The following hyperparameters were used during training: |
|
|
|
- learning_rate: 1E+00 |
|
- per device batch size: 60 |
|
- effective batch size: 180 |
|
- seed: 42 |
|
- optimizer: AdamW with betas (0.9,0.999) and eps 1E-08 |
|
- weight decay: 2E-02 |
|
- D-Adaptation: True |
|
- Warmup: True |
|
- number of epochs: 15 |
|
- mixed_precision_training: bf16 |
|
|
|
## Training results |
|
| Epoch | Train Loss | Validation Loss | |
|
| ----- | ---------- | --------------- | |
|
| 0 | 0.0554 | 0.047 | |
|
| 1 | 0.044 | 0.0472 | |
|
| 2 | 0.0374 | 0.0425 | |
|
| 3 | 0.0322 | 0.041 | |
|
| 4 | 0.0278 | 0.0403 | |
|
| 5 | 0.0246 | 0.0389 | |
|
| 6 | 0.0215 | 0.0389 | |
|
| 7 | 0.0192 | 0.0388 | |
|
| 8 | 0.017 | 0.0379 | |
|
| 9 | 0.0154 | 0.0375 | |
|
| 10 | 0.0142 | 0.0381 | |
|
| 11 | 0.0132 | 0.0372 | |
|
| 12 | 0.0126 | 0.0377 | |
|
| 13 | 0.012 | 0.0377 | |
|
|
|
## Evaluation results |
|
| Epoch | top_1 | top_3 | top_5 | top_10 | top_25 | |
|
| ----- | ----- | ----- | ----- | ----- | ----- | |
|
| 0 | 0.373 | 0.476 | 0.509 | 0.544 | 0.573 | |
|
| 1 | 0.362 | 0.466 | 0.501 | 0.537 | 0.568 | |
|
| 2 | 0.371 | 0.476 | 0.511 | 0.546 | 0.576 | |
|
| 3 | 0.369 | 0.473 | 0.506 | 0.54 | 0.569 | |
|
| 4 | 0.373 | 0.478 | 0.512 | 0.547 | 0.578 | |
|
| 5 | 0.378 | 0.483 | 0.517 | 0.552 | 0.58 | |
|
| 6 | 0.371 | 0.475 | 0.509 | 0.543 | 0.571 | |
|
| 7 | 0.379 | 0.484 | 0.517 | 0.55 | 0.578 | |
|
| 8 | 0.378 | 0.482 | 0.515 | 0.548 | 0.575 | |
|
| 9 | 0.383 | 0.489 | 0.523 | 0.556 | 0.584 | |
|
| 10 | 0.38 | 0.483 | 0.517 | 0.549 | 0.575 | |
|
| 11 | 0.38 | 0.485 | 0.518 | 0.551 | 0.577 | |
|
| 12 | 0.383 | 0.489 | 0.522 | 0.556 | 0.582 | |
|
| 13 | 0.385 | 0.49 | 0.523 | 0.555 | 0.581 | |
|
|
|
## Framework versions |
|
- Transformers: 4.25.1 |
|
- PyTorch: 2.0.0.dev20230210+cu118 |
|
- PyTorch Lightning: 1.8.6 |
|
- Datasets: 2.7.1 |
|
- Tokenizers: 0.13.1 |
|
- Sentence Transformers: 2.2.2 |
|
|
|
## Additional Information |
|
This model was trained as part of my Master's Thesis **'Evaluation of transformer based language models for use in service information systems'**. The source code is available on [Github](https://github.com/LLukas22/Retrieval-Augmented-QA). |
