multimodalart's picture
Squashing commit
4450790 verified
raw
history blame
2.49 kB
import torch
def calc_mantissa(abs_x, exponent, normal_mask, MANTISSA_BITS, EXPONENT_BIAS, generator=None):
mantissa_scaled = torch.where(
normal_mask,
(abs_x / (2.0 ** (exponent - EXPONENT_BIAS)) - 1.0) * (2**MANTISSA_BITS),
(abs_x / (2.0 ** (-EXPONENT_BIAS + 1 - MANTISSA_BITS)))
)
mantissa_scaled += torch.rand(mantissa_scaled.size(), dtype=mantissa_scaled.dtype, layout=mantissa_scaled.layout, device=mantissa_scaled.device, generator=generator)
return mantissa_scaled.floor() / (2**MANTISSA_BITS)
#Not 100% sure about this
def manual_stochastic_round_to_float8(x, dtype, generator=None):
if dtype == torch.float8_e4m3fn:
EXPONENT_BITS, MANTISSA_BITS, EXPONENT_BIAS = 4, 3, 7
elif dtype == torch.float8_e5m2:
EXPONENT_BITS, MANTISSA_BITS, EXPONENT_BIAS = 5, 2, 15
else:
raise ValueError("Unsupported dtype")
x = x.half()
sign = torch.sign(x)
abs_x = x.abs()
sign = torch.where(abs_x == 0, 0, sign)
# Combine exponent calculation and clamping
exponent = torch.clamp(
torch.floor(torch.log2(abs_x)) + EXPONENT_BIAS,
0, 2**EXPONENT_BITS - 1
)
# Combine mantissa calculation and rounding
normal_mask = ~(exponent == 0)
abs_x[:] = calc_mantissa(abs_x, exponent, normal_mask, MANTISSA_BITS, EXPONENT_BIAS, generator=generator)
sign *= torch.where(
normal_mask,
(2.0 ** (exponent - EXPONENT_BIAS)) * (1.0 + abs_x),
(2.0 ** (-EXPONENT_BIAS + 1)) * abs_x
)
inf = torch.finfo(dtype)
torch.clamp(sign, min=inf.min, max=inf.max, out=sign)
return sign
def stochastic_rounding(value, dtype, seed=0):
if dtype == torch.float32:
return value.to(dtype=torch.float32)
if dtype == torch.float16:
return value.to(dtype=torch.float16)
if dtype == torch.bfloat16:
return value.to(dtype=torch.bfloat16)
if dtype == torch.float8_e4m3fn or dtype == torch.float8_e5m2:
generator = torch.Generator(device=value.device)
generator.manual_seed(seed)
output = torch.empty_like(value, dtype=dtype)
num_slices = max(1, (value.numel() / (4096 * 4096)))
slice_size = max(1, round(value.shape[0] / num_slices))
for i in range(0, value.shape[0], slice_size):
output[i:i+slice_size].copy_(manual_stochastic_round_to_float8(value[i:i+slice_size], dtype, generator=generator))
return output
return value.to(dtype=dtype)