tencent/Youtu-Parsing
Full fine-tune (fewer layers) vs tencent/Youtu-LLM-2B. tencent/Youtu-LLM-2B matches part of tencent/Youtu-Parsing (385 of its 827 tensors; the rest of tencent/Youtu-Parsing has no counterpart). On the shared part: Fine-tune: 100% of the weights changed, across 32 of 32 layers.
Tensors changed: Norms 129/129, MLP 96/96, Attention 160/160
1 other repos hold byte-for-byte identical weight files.