Menu

BF16 FP8: checking precision across two CPU compiler versions
πŸ“°

BF16 FP8: checking precision across two CPU compiler versions

DEV Community: pytorchΒ·BluccaΒ·8 days ago
#7Xf9Lnwi
#devtopytorch#devto#dev#torch#fp32
Reading 0:00
15s threshold

A dtype conversion can remain visible after the final FP8 quantization. It belongs in a fusion's numerical contract, even when its intermediate allocation is removed.

Here is a small CPU example:

x = torch . tensor ([ 1.0626 ], dtype = torch . float32 ) x . to ( torch . bfloat16 ). to ( torch . float8_e4m3fn ). float () # 1.0 x . to ( torch . float8_e4m3fn ). float () # 1.125

Enter fullscreen mode

Exit fullscreen mode

The exact FP32 input is 1.062600016593933 . BF16 rounds it to 1.0625 : the midpoint between the two FP8 values 1.0 and 1.125 . Round-to-nearest-even sends that midpoint to 1.0 ; the original FP32 value sits just above it.

A complete control band

I used all 65,536 exactly representable FP32 values 1 + i/65536 , for i = 0…65535 . The two eager conversion paths differ at 2,048 values, or 3.125% . A standard-library checker reproduces the geometry.

For CPU compilation I tested these two functions:

def cast_only ( x ): return x . to ( torch . bfloat16 ). to ( torch . float8_e4m3fn ).…

Continue reading β€” create a free account

Join HashtagPLUS to read full articles, follow hashtags, vote, and join the conversation.