CogniSoftOrg/canary-1b-v2-mlx-bf16
Same weights, converted vs nvidia/canary-1b-v2. Same weights stored in another dtype (F32/I64 → BF16), plus 708 new tensors.
Tensors changed: other 0/192, MLP 0/256, Norms 0/320
Same weights, converted vs nvidia/canary-1b-v2. Same weights stored in another dtype (F32/I64 → BF16), plus 708 new tensors.
Tensors changed: other 0/192, MLP 0/256, Norms 0/320