soichisumi
/

bge-reranker-v2-m3-mlx

Model card Files Files and versions

bge-reranker-v2-m3 — MLX (bf16)

BAAI/bge-reranker-v2-m3 の MLX 変換 (bfloat16、量子化なし)。1.1 GB。

変換: mlx_lm convert --hf-path BAAI/bge-reranker-v2-m3 --mlx-path . --dtype bfloat16 (bge-m3 用 MLX 互換 patch を適用)
評価: mteb/scidocs-reranking 30 クエリ / 897 ペア
結果: Kendall τ = 0.9954, nDCG@10 Δ = 0.0000 → PASS+ (lossless)

Downloads last month: 63

Safetensors

Model size

0.6B params

Tensor type

BF16

·

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for soichisumi/bge-reranker-v2-m3-mlx

Base model

BAAI/bge-reranker-v2-m3

Finetuned

(78)

this model