wolfram/miquliz-120b-v2.0
Partial fine-tune (fewer layers) vs 152334H/miqu-1-70b-sf. 152334H/miqu-1-70b-sf matches part of wolfram/miquliz-120b-v2.0 (723 of its 1263 tensors; the rest of wolfram/miquliz-120b-v2.0 has no counterpart). On the shared part: Partial change: 79% of the weights changed (attention, lm_head, mlp, norms), in 73 of 80 layers.
Tensors changed: Embeddings 0/1, Norms 120/161, MLP 193/240, Attention 253/320, lm_head 1/1
2 derivatives checked
- mradermacher/miquliz-120b-v2.0-GGUF: Quantization
- mradermacher/miquliz-120b-v2.0-i1-GGUF: Quantization
1 other repos hold byte-for-byte identical weight files.