Qwen/Qwen3Guard-Stream-4B

Full fine-tune vs Qwen/Qwen3-4B. Fine-tune: 90% of the weights changed, across 36 of 36 layers.

Tensors changed: Embeddings 0/1, Norms 129/145, MLP 108/108, Attention 144/144

2 derivatives checked

• geoffmunn/Qwen3Guard-Stream-4B: Quantization
• zenlm/zen-guard-stream-4b: Identical copy

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

On the Hub · Downloads on Model Pulse