mradermacher/OpenELM-450M-GGUF
Same weights, converted vs apple/OpenELM-450M. Same weights stored in another dtype (F32 → F16/F32), plus 40 new tensors.
Tensors changed: Norms 0/81, Attention 0/40, MLP 0/20, Embeddings 0/1
Same weights, converted vs apple/OpenELM-450M. Same weights stored in another dtype (F32 → F16/F32), plus 40 new tensors.
Tensors changed: Norms 0/81, Attention 0/40, MLP 0/20, Embeddings 0/1