README.md

---
license: apache-2.0
language:
- en
library_name: transformers
pipeline_tag: sentence-similarity
tags:
- sentence-transformers
- feature-extraction
- ONNX
- Optimum
- ONNXRuntime
inference: false

---
# ONNX version of `sentence-transformers/all-mpnet-base-v2`

This is the ONNX version of https://huggingface.co/sentence-transformers/all-mpnet-base-v2, examined that the produced embeddings are the same.

Optmized for CPU usage.

## Convert

The same checkpoint can also be created by using the `convert.py` script.

## Usage - `transformers`

Exactly the same as in `sentence-transformers/all-mpnet-base-v2` except using `ORTModelForFeatureExtraction` from optimum.

```bash
pip install optimum[onnxruntime]
```

```python
from transformers import AutoTokenizer
from optimum.onnxruntime import ORTModelForFeatureExtraction
import torch
import torch.nn.functional as F

# Mean Pooling - Take attention mask into account for correct averaging
def mean_pooling(model_output, attention_mask):
    token_embeddings = model_output[0] #First element of model_output contains all token embeddings
    input_mask_expanded = attention_mask.unsqueeze(-1).expand(token_embeddings.size()).float()
    return torch.sum(token_embeddings * input_mask_expanded, 1) / torch.clamp(input_mask_expanded.sum(1), min=1e-9)


# Sentences we want sentence embeddings for
sentences = ['This is an example sentence', 'Each sentence is converted']

# Load model from HuggingFace Hub
tokenizer = AutoTokenizer.from_pretrained('yilunzhang/all-mpnet-base-v2-onnx')
model = ORTModelForFeatureExtraction.from_pretrained('yilunzhang/all-mpnet-base-v2-onnx')

# Tokenize sentences
encoded_input = tokenizer(sentences, padding=True, truncation=True, return_tensors='pt')

# Compute token embeddings
with torch.no_grad():
    model_output = model(**encoded_input)

# Perform pooling
sentence_embeddings = mean_pooling(model_output, encoded_input['attention_mask'])

# Normalize embeddings
sentence_embeddings = F.normalize(sentence_embeddings, p=2, dim=1)

print("Sentence embeddings:")
print(sentence_embeddings)
```