File size: 1,790 Bytes
179528d
 
0bbe73f
 
4815a23
 
179528d
1eb4f7b
 
40ce004
 
105e839
40ce004
3445ad5
40ce004
3445ad5
40ce004
3445ad5
40ce004
3445ad5
105e839
3445ad5
105e839
4815a23
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
---
license: mit
language:
- en
tags:
- sparse sparsity quantized onnx embeddings int8
---
This is the sparsified ONNX variant of the [bge-large-en-v1.5](https://huggingface.co/BAAI/bge-large-en-v1.5) embeddings model created with [DeepSparse Optimum](https://github.com/neuralmagic/optimum-deepsparse) for ONNX export/inference pipeline and Neural Magic's [Sparsify](https://account.neuralmagic.com/signin?client_id=d04a5f0c-983d-11ed-88a6-971073f187d3&return_to=https%3A//accounts.neuralmagic.com/v1/connect/authorize%3Fscope%3Dsparsify%3Aread%2Bsparsify%3Awrite%2Buser%3Aapi-key%3Aread%2Buser%3Aprofile%3Awrite%2Buser%3Aprofile%3Aread%26response_type%3Dcode%26code_challenge_method%3DS256%26redirect_uri%3Dhttps%3A//apps.neuralmagic.com/sparsify/oidc/callback.html%26state%3Da9b466a6193c4a7b92cba469408d2495%26client_id%3Dd04a5f0c-983d-11ed-88a6-971073f187d3%26code_challenge%3DP0EkmKBpplTb7crJOGS8YLSwT8UH-BeuD0wuE4JTORQ%26response_mode%3Dquery) for One-Shot INT8 quantization and unstructured pruning (50%). 

Model achieves 98% accuracy recovery on the STSB validation dataset vs. [dense ONNX variant](https://huggingface.co/zeroshot/bge-base-en-v1.5-dense).

Current up-to-date list of sparse and quantized bge ONNX models:

[zeroshot/bge-large-en-v1.5-sparse](https://huggingface.co/zeroshot/bge-large-en-v1.5-sparse)

[zeroshot/bge-large-en-v1.5-quant](https://huggingface.co/zeroshot/bge-large-en-v1.5-quant)

[zeroshot/bge-base-en-v1.5-sparse](https://huggingface.co/zeroshot/bge-base-en-v1.5-sparse)

[zeroshot/bge-base-en-v1.5-quant](https://huggingface.co/zeroshot/bge-base-en-v1.5-quant)

[zeroshot/bge-small-en-v1.5-sparse](https://huggingface.co/zeroshot/bge-small-en-v1.5-sparse)

[zeroshot/bge-small-en-v1.5-quant](https://huggingface.co/zeroshot/bge-small-en-v1.5-quant)