intfloat/multilingual-e5-small request

[bm777]

Request of intfloat/multilingual-e5-small. That would be nice :)

[ashishu23]

Hi @bm777, I am interested in solving this issue. Wanted to confirm it the issue still persists?
If not, then can I work on this?

[bm777]

@ashishu23 go on.

[ashishu23]

Hi @bm777 , can you please tell whether the issue is solved in the last pull request?
If not then what else could be done?

[andersfylling]

@ashishu23 which pr?

[bm777]

Sorry for the late reply @ashishu23 .
i finally decided to convert the model into onnx format and quantized it into qint8 using the latest version of the sentence-transformers library -> here

[joein]

this model can be added to fastembed v0.6.0 in runtime

it requires the following code:

from fastembed import TextEmbedding
from fastembed.common.model_description import PoolingType, ModelSource

TextEmbedding.add_custom_model(
    model="intfloat/multilingual-e5-small",
    pooling=PoolingType.MEAN,
    normalization=True,
    sources=ModelSource(hf="intfloat/multilingual-e5-small"),  # can be used with an `url` to load files from a private storage
    dim=384,
    model_file="onnx/model.onnx",  # can be used to load an already supported model with another optimization or quantization, e.g. onnx/model_O4.onnx
)
model = TextEmbedding(model_name="intfloat/multilingual-e5-small")
embeddings = list(model.embed(documents))