Mihaiii/Covasna-0.1
Partial fine-tune (fewer layers) vs migtissera/Tess-70B-v1.6. Mihaiii/Covasna-0.1 matches part of migtissera/Tess-70B-v1.6 (435 of its 723 tensors; the rest of migtissera/Tess-70B-v1.6 has no counterpart). On the shared part: Partial change: 84% of the weights changed (attention, mlp, norms), in 41 of 48 layers.
Tensors changed: Embeddings 0/1, Norms 82/97, MLP 123/144, Attention 164/192, lm_head 0/1
1 derivatives checked
- mradermacher/Covasna-0.1-GGUF: Quantization