mradermacher/Fijik1.5-2.6b-a380m-GGUF
Same weights, converted vs Pinkstack/Fijik1.5-2.6b-a380m. Same weights stored in another dtype (BF16 → F16/F32), plus 90 new tensors.
Tensors changed: Embeddings 0/1, other 0/30, Norms 0/61, Attention 0/120
Same weights, converted vs Pinkstack/Fijik1.5-2.6b-a380m. Same weights stored in another dtype (BF16 → F16/F32), plus 90 new tensors.
Tensors changed: Embeddings 0/1, other 0/30, Norms 0/61, Attention 0/120