DazzlingXeno/MS-Beast-33b
Partial fine-tune (fewer layers) vs DazzlingXeno/MS-Drummer-Sunfall-22b. DazzlingXeno/MS-Drummer-Sunfall-22b matches part of DazzlingXeno/MS-Beast-33b (507 of its 759 tensors; the rest of DazzlingXeno/MS-Beast-33b has no counterpart). On the shared part: Partial change: 25% of the weights changed (attention, mlp, norms), in 14 of 56 layers.
Tensors changed: lm_head 0/1, Embeddings 0/1, Norms 28/113, MLP 42/168, Attention 56/224
2 derivatives checked
- mradermacher/MS-Beast-33b-i1-GGUF: Quantization
- mradermacher/MS-Beast-33b-GGUF: Quantization