[Questions] Exact Parameter Counts & Minimum GPU Vram Requirements
#11
by wjkim9653 - opened
The model card says this model has size of 71B params, and each params are in NVFP4 format.
I wonder if this is actually accurate. If true, then the minimum vram requirement to run this model would be roughly 30~40GBs which I doubt.
Isn't it supposed to be 122B parameters in NVFP4 format? (w/ minimum vram requirement at around 70GBs?)
I'm not familiar with NVFP4, so I might be wrong. If so, please let me know.