Qwen/Qwen-Image-Layered

Full fine-tune vs Qwen/Qwen-Image. Fine-tune: 99% of the weights changed, across 0 of 0 layers.

Tensors changed: other 244/246, Embeddings 4/4, Attention 957/960, Norms 242/243, MLP 476/480

3 derivatives checked

• unsloth/Qwen-Image-Layered-GGUF: Same weights, converted
• zimengxiong/Qwen-Image-Layered-6bit: Quantization
• T5B/Qwen-Image-Layered-FP8: Full fine-tune

On the Hub · Downloads on Model Pulse