YAML Metadata Warning:empty or missing yaml metadata in repo card

Check out the documentation for more information.

ModernBERT-base (MLX)

MLX-compatible ModernBERT-base encoder weights for Apple Silicon.

Usage with autoresearch-embed

Architecture

  • 22-layer encoder, 768 hidden, 12 heads
  • Alternating local (128-window) and global attention
  • GLU MLP with GeLU activation
  • RoPE positional encodings (local theta=10000, global theta=160000)
  • 149.2M parameters

Source

Weights from answerdotai/ModernBERT-base, packaged for direct loading into MLX nn.Module via safetensors.

Framework

  • MLX >= 0.30.0
  • Apple Silicon (M1/M2/M3/M4)
Downloads last month
2
Safetensors
Model size
0.1B params
Tensor type
F32
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support