bytedance-research/ATI
Full fine-tune vs Wan-AI/Wan2.1-I2V-14B-480P. Fine-tune: 100% of the weights changed, across 40 of 40 layers.
Tensors changed: Attention 800/800, Norms 280/280, MLP 160/160, other 53/53, Embeddings 10/10
Full fine-tune vs Wan-AI/Wan2.1-I2V-14B-480P. Fine-tune: 100% of the weights changed, across 40 of 40 layers.
Tensors changed: Attention 800/800, Norms 280/280, MLP 160/160, other 53/53, Embeddings 10/10