ehristoforu/BigFalcon3-36B
Partial fine-tune (fewer layers) vs tiiuae/Falcon3-10B-Instruct. tiiuae/Falcon3-10B-Instruct matches part of ehristoforu/BigFalcon3-36B (363 of its 651 tensors; the rest of ehristoforu/BigFalcon3-36B has no counterpart). On the shared part: Partial change: 74% of the weights changed (attention, mlp, norms), in 32 of 40 layers.
Tensors changed: Embeddings 0/1, Norms 64/81, MLP 96/120, Attention 128/160, lm_head 0/1
2 derivatives checked
- mradermacher/BigFalcon3-36B-i1-GGUF: Quantization
- mradermacher/BigFalcon3-36B-GGUF: Quantization
1 other repos hold byte-for-byte identical weight files.