xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-03-25 00:29:13 +08:00
Code Issues Packages Projects Releases Wiki Activity

[Misc] Fix ScalarType float4 naming (#17690)

Browse Source 
Signed-off-by: Lucas Wilkinson <lwilkinson@neuralmagic.com>
This commit is contained in:
Lucas Wilkinson 2025-05-06 04:07:15 -04:00 committed by GitHub
parent 63ced7b43f
commit 6eae34533a
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
3 changed files with 3 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
tests/kernels/quantization/test_nvfp4_quant.py
View File

@ -17,7 +17,7 @@ PAD_SHAPES = [(90, 64), (150, 64), (128, 48), (128, 80), (150, 80), (90, 48),
SEEDS = [42]
CUDA_DEVICES = ['cuda:0']
FLOAT4_E2M1_MAX = scalar_types.float4_e2m1fn.max()
FLOAT4_E2M1_MAX = scalar_types.float4_e2m1f.max()
FLOAT8_E4M3_MAX = torch.finfo(torch.float8_e4m3fn).max
# E2M1 to float

2
tests/test_scalartype.py
View File

@ -11,7 +11,7 @@ from vllm.scalar_type import scalar_types
(0, 15, scalar_types.uint4),
(-8, 7, scalar_types.uint4b8),
(-128, 127, scalar_types.uint8b128),
(-6., 6., scalar_types.float4_e2m1fn),
(-6., 6., scalar_types.float4_e2m1f),
(-28., 28., scalar_types.float6_e3m2f),
(torch.int8, scalar_types.int8),
(torch.uint8, scalar_types.uint8),

2
vllm/scalar_type.py
View File

@ -333,7 +333,7 @@ class scalar_types:
float6_e3m2f = ScalarType.float_(3, 2, True, NanRepr.NONE)
# fp4, https://www.opencompute.org/documents/ocp-microscaling-formats-mx-v1-0-spec-final-pdf
float4_e2m1fn = ScalarType.float_(2, 1, True, NanRepr.NONE)
float4_e2m1f = ScalarType.float_(2, 1, True, NanRepr.NONE)
# "gptq" types
uint2b2 = ScalarType.uint(2, 2)