mlx-community/openai-privacy-filter-bf16
Same weights, converted vs openai/privacy-filter. Same weights stored in another dtype (BF16/F32 → BF16), plus 48 new tensors.
Tensors changed: Attention 0/72, Embeddings 0/1, Norms 0/17, MLP 0/16, other 0/2
Same weights, converted vs openai/privacy-filter. Same weights stored in another dtype (BF16/F32 → BF16), plus 48 new tensors.
Tensors changed: Attention 0/72, Embeddings 0/1, Norms 0/17, MLP 0/16, other 0/2