ByteDance-Seed/BAGEL-7B-MoT
Full fine-tune (fewer layers) vs Qwen/Qwen2.5-7B-Instruct. Qwen/Qwen2.5-7B-Instruct matches part of ByteDance-Seed/BAGEL-7B-MoT (339 of its 1467 tensors; the rest of ByteDance-Seed/BAGEL-7B-MoT has no counterpart). On the shared part: Fine-tune: 100% of the weights changed, across 28 of 28 layers.
Tensors changed: Embeddings 1/1, Norms 57/57, MLP 84/84, Attention 196/196, lm_head 1/1
1 derivatives checked
- calcuis/bagel-gguf: Same weights, repacked