F16 model size is same as the original F32 one?
#3
by
KSimulation
- opened
F16 model size is same as the original F32 one?
[23.8G]
Am I missing anything...
The original is in BF16 not FP32, so this is expected.
The FP16 GGUF is meant as a base for other (for exampke _K) quants via CPP code, although they're all currently created from the safetensors file.
city96
changed discussion status to
closed