Typesense Public Embedding Models
We store our current supported embedding models in this repo and you can also convert your own models to ONNX format and create a PR to add it to our supported models list.
Convert a model to ONNX format
Converting a Hugging Face Transformers Model
You can follow instructions from this link to convert any model from Hugging Face to ONNX format using optimum-cli.
Converting a PyTorch Model
You can use torch.onnx APIs to convert PyTorch models to ONNX.
Converting a Tensorflow Model
You can use tf2onnx tool to convert Tensorflow models to ONNX.
Creating model config
Before creating a PR with your ONNX model, you should store model file, vocab file and model config file under a folder with model name. Your model config must be named as config.json and should contain those keys:
| Key | Description | Optional |
|---|---|---|
| model_md5 | MD5 checksum of model file as string | No |
| vocab_md5 | MD5 checksum of vocab file as string | No |
| model_type | Model type (currently only bert and xlm_roberta supported) |
No |
| vocab_file_name | File name of vocab file | No |
| indexing_prefix | Prefix to be added before embedding documents | Yes |
| query_prefix | Prefix to be added before embedding queries | Yes |