Downloads · 30 days
111
6% of all-time downloads
bge-m3-onnx-int8
raludi/bge-m3-onnx-int8
bge-m3-onnx-int8 is a machine learning model from raludi. Use it for the machine learning task on the model card, and read the license before you ship it in a product. The card lists the license as mit.
Quantized from the official BAAI/bge-m3 onnx/model.onnx (fp32, 2.3GB) with onnxruntime.quantization.quantizedynamic (QInt8, perchannel=True, MatMulConstBOnly=True). 570MB, drop-in for text-embeddings-inference CPU ima…
All-time downloads
1.7K
Public
Repo size
587 MB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.onnx570 MB · 97%
At a glance
- License
- mit
- Model type
- xlm-roberta
- Access
- Public
- Created
- Jul 10, 2026
- Updated
- Jul 10, 2026
- SHA
- 00f2ae1a
Base models
Model card
From the Hugging Face model README
BGE-M3 — ONNX int8 (dynamic, per-channel)
Quantized from the official BAAI/bge-m3
onnx/model.onnx (fp32, 2.3GB) with onnxruntime.quantization.quantize_dynamic
(QInt8, per_channel=True, MatMulConstBOnly=True). 570MB, drop-in for
text-embeddings-inference
CPU images (--model-id raludi/bge-m3-onnx-int8). Dense 1024-dim output.
Validation vs fp32 (12 Spanish business documents + prototype classification): mean cosine 0.989 (min 0.986), nearest-neighbour flips only on technical ties (top1-top2 margin < 0.007), prototype classification agreement 3/3.
Quantized with onnxruntime 1.27.0. Same license as the base model (MIT).
- Type
- xlm-roberta
- License
- mit
- Created
- Jul 10, 2026
- Updated
- Jul 10, 2026