Downloads · 30 days
13
25% of all-time downloads
maccotaro0/japanese-reranker-xsmall-v2-onnx
japanese-reranker-xsmall-v2-onnx is a text classification model from maccotaro0. Use it when you need a label for a piece of text. The card lists the license as mit.
This is an ONNX conversion of hotchpotch/japanese-reranker-xsmall-v2 (MIT License, © hotchpotch). Weights are unchanged except for dynamic int8 quantization.
Downloads · 30 days
13
25% of all-time downloads
All-time downloads
52
Public
Repo size
27.4 MB
Likes
0
Public
Click a slice to open those files.
.gz27.4 MB · 80%
From the Hugging Face model README
This is an ONNX conversion of hotchpotch/japanese-reranker-xsmall-v2 (MIT License, © hotchpotch). Weights are unchanged except for dynamic int8 quantization.
onnx/model_quantized.onnx — exported with optimum-cli export onnx --task text-classification, then
onnxruntime.quantization.quantize_dynamic(weight_type=QInt8) (147 MB → 37 MB).input_ids, attention_mask. Output: logits (one score per query–passage pair; higher is more relevant).Used by exista personal as an on-device reranker (onnxruntime-node / transformers.js). Measured on Apple Silicon (CPU, int8, 10 pairs): 128 tokens ≈ 63 ms, 192 tokens ≈ 106 ms.
MIT License, same as the original model. See the original repository for the full text and attribution.