Divij/Qwen3.5-4B-SFT-decompiler-O0-full

Full fine-tune vs Qwen/Qwen3.5-4B. Fine-tune: 90% of the weights changed, across 32 of 32 layers.

Tensors changed: Embeddings 1/5, MLP 96/199, Attention 190/324, Norms 71/209, other 0/1

1 derivatives checked

• sillykiwi/Qwen3.5-4B-SFT-decompiler-O0-full-Q4_K_M-GGUF: Quantization

On the Hub · Downloads on Model Pulse