jsfs11/TemptressTensor-10.7B-v0.1a
Partial fine-tune (fewer layers) vs jsfs11/RandomMergeNoNormWEIGHTED-7B-DARETIES. jsfs11/RandomMergeNoNormWEIGHTED-7B-DARETIES matches part of jsfs11/TemptressTensor-10.7B-v0.1a (291 of its 435 tensors; the rest of jsfs11/TemptressTensor-10.7B-v0.1a has no counterpart). On the shared part: Partial change: 69% of the weights changed (attention, mlp, norms), in 23 of 32 layers.
Tensors changed: MLP 69/96, Attention 92/128, Norms 46/65, Embeddings 0/1, lm_head 0/1
1 derivatives checked
- mradermacher/TemptressTensor-10.7B-v0.1a-GGUF: Quantization