Gå offline med appen Player FM !
Half precision
Manage episode 301973966 series 2921809
In this episode I talk about reduced precision floating point formats float16 (aka half precision) and bfloat16. I'll discuss what floating point numbers are, how these two formats vary, and some of the practical considerations that arise when you are working with numeric code in PyTorch that also needs to work in reduced precision. Did you know that we do all CUDA computations in float32, even if the source tensors are stored as float16? Now you know!
Further reading.
- The Wikipedia article on IEEE floating point is pretty great https://en.wikipedia.org/wiki/IEEE_754
- How bfloat16 works out when doing training https://arxiv.org/abs/1905.12322
- Definition of acc_type in PyTorch https://github.com/pytorch/pytorch/blob/master/aten/src/ATen/AccumulateType.h
82 episoder
Manage episode 301973966 series 2921809
In this episode I talk about reduced precision floating point formats float16 (aka half precision) and bfloat16. I'll discuss what floating point numbers are, how these two formats vary, and some of the practical considerations that arise when you are working with numeric code in PyTorch that also needs to work in reduced precision. Did you know that we do all CUDA computations in float32, even if the source tensors are stored as float16? Now you know!
Further reading.
- The Wikipedia article on IEEE floating point is pretty great https://en.wikipedia.org/wiki/IEEE_754
- How bfloat16 works out when doing training https://arxiv.org/abs/1905.12322
- Definition of acc_type in PyTorch https://github.com/pytorch/pytorch/blob/master/aten/src/ATen/AccumulateType.h
82 episoder
Alla avsnitt
×Välkommen till Player FM
Player FM scannar webben för högkvalitativa podcasts för dig att njuta av nu direkt. Den är den bästa podcast-appen och den fungerar med Android, Iphone och webben. Bli medlem för att synka prenumerationer mellan enheter.