Marketplace Model: Small Multilingual PII Masking

From Minibase

Small Multilingual PII Masking

Identify and mask PII with low latency with this multilingual PII masking model. This model was trained on 99k examples that allow for high accuracy masking.

Basic Information

Base Model:Small Base
Created by:Michaelminibase
Times imported:822
Released:Sep 24, 2025
Model Size:138 MB
Model Type:Causal Language Model
Format:HIGH

Technical Details

Hidden Size:576
Hidden Layers:30
Attention Heads:9
Vocabulary Size:49,152
Max Context Length:2,048 tokens
Precision:BFloat16 (BF16)
Learning Rate:0.000050
Training Epochs:3
Effective Batch Size:16
Optimizer:AdamW

Training Datasets

NameTypeExamplesSize
Multilingual PII Masking (Part 12)SFT10,0004.9 MB
Multilingual PII Masking (Part 13)SFT10,0004.8 MB
Multilingual PII Masking (Part 14)SFT10,0004.8 MB
Multilingual PII Masking (Part 15)SFT10,0004.8 MB
Multilingual PII Masking (Part 16)SFT10,0004.9 MB
Multilingual PII Masking (Part 17)SFT10,0004.9 MB
Multilingual PII Masking (Part 18)SFT10,0004.8 MB
Multilingual PII Masking (Part 19)SFT10,0004.8 MB
Multilingual PII Masking (Part 20)SFT10,0004.8 MB
Multilingual PII Masking (Part 21)SFT9,2614.5 MB