DevQuasar-8/ibm-granite.granite-3.0-3b-a800m-base-GGUF

Same weights, converted vs ibm-granite/granite-3.0-3b-a800m-base. Same weights stored in another dtype (F32 → F16/F32), plus 64 new tensors.

Tensors changed: lm_head 0/1, Embeddings 0/1, other 0/64, Norms 0/65, Attention 0/128

On the Hub · Downloads on Model Pulse