biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[Quantization] Add compressed-tensors NVFP4 MoE Support (#19990)

Browse Source 
Signed-off-by: Dipika Sikka <dipikasikka1@gmail.com>
Signed-off-by: Dipika <dipikasikka1@gmail.com>
This commit is contained in:
Dipika Sikka
2025-06-30 00:05:40 +02:00
committed by GitHub
parent 7b1895e6ce
commit 6f2f53a82d
6 changed files with 295 additions and 22 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
tests/quantization/test_compressed_tensors.py
View File

@@ -17,7 +17,7 @@ from vllm.model_executor.layers.quantization.compressed_tensors.compressed_tenso
CompressedTensorsW4A4Fp4, CompressedTensorsW4A16Fp4,
CompressedTensorsW4A16Sparse24, CompressedTensorsW8A8Fp8,
CompressedTensorsW8A8Int8, CompressedTensorsW8A16Fp8,
CompressedTensorsWNA16)
CompressedTensorsWNA16, cutlass_fp4_supported)
from vllm.model_executor.layers.quantization.utils.w8a8_utils import (
sparse_cutlass_supported)
from vllm.platforms import current_platform
@@ -668,8 +668,8 @@ def test_compressed_tensors_nvfp4(vllm_runner, args):
assert isinstance(qkv_proj.quant_method,
CompressedTensorsLinearMethod)
if isinstance(qkv_proj.scheme, scheme) or isinstance(
qkv_proj.scheme, CompressedTensorsW4A16Fp4
) and not CompressedTensorsW4A4Fp4.cutlass_fp4_supported():
qkv_proj.scheme,
CompressedTensorsW4A16Fp4) and not cutlass_fp4_supported():
assert True
else:
raise AssertionError("FP4 Scheme Mismatch")