OpenMed/privacy-filter-nemotron

Partial fine-tune vs openai/privacy-filter. Partial change: 10% of the weights changed (attention, embeddings, mlp, norms), in 8 of 8 layers.

Tensors changed: Attention 72/72, Embeddings 1/1, Norms 17/17, MLP 16/48

2 derivatives checked

• LocalAI-io/privacy-filter-nemotron-GGUF: Same weights, converted
• OpenMed/privacy-filter-nemotron-mlx-8bit: Quantization

2 other repos hold byte-for-byte identical weight files.

On the Hub · Downloads on Model Pulse