microsoft

/

phi-2

like 3.19k

@a8nova The pytorch weights are sharded and converted into safetensors for reducing the VRAM required for loading the model.
You can find them in the model's files with names similar to model-00001-of-00002.safetensors

Thanks for the answer @DeathReaper0965 !

Hi @DeathReaper0965 - thank you for response. I wanted to create a libtorchlite version of the weights, and my understanding is that I need the pytorch weights to be able to do that, can i create a libtorchlite version of this model using the safetensors?