EngineX/xc-llm-ascend
3
0
Fork 0
Code Issues Pull Requests Actions Projects Releases Wiki Activity

fix nz for quantization (#4943)

Browse Source 
quantization ops rely on NZ by force, we should remove the nz check for it.

Signed-off-by: wangxiyuan <wangxiyuan1007@gmail.com>
This commit is contained in:
wangxiyuan
2025-12-12 14:54:41 +08:00
committed by GitHub
parent 5932abc446
commit 01a13a9b77
4 changed files with 11 additions and 14 deletions
Download Patch File Download Diff File Expand all files Collapse all files

11
vllm_ascend/quantization/w4a8_dynamic.py
View File

@@ -27,7 +27,7 @@ from vllm.forward_context import get_forward_context
from vllm_ascend.ascend_config import get_ascend_config
from vllm_ascend.distributed.parallel_state import get_mc2_group
from vllm_ascend.ops.moe.experts_selector import select_experts
from vllm_ascend.utils import ACL_FORMAT_FRACTAL_NZ, is_enable_nz
from vllm_ascend.utils import ACL_FORMAT_FRACTAL_NZ
class AscendW4A8DynamicLinearMethod:
@@ -482,10 +482,9 @@ class AscendW4A8DynamicFusedMoEMethod:
self.update_bias(layer, w13_bias, w2_bias)
if is_enable_nz():
layer.w13_weight.data = torch_npu.npu_format_cast(
layer.w13_weight.data, ACL_FORMAT_FRACTAL_NZ)
layer.w2_weight.data = torch_npu.npu_format_cast(
layer.w2_weight.data, ACL_FORMAT_FRACTAL_NZ)
layer.w13_weight.data = torch_npu.npu_format_cast(
layer.w13_weight.data, ACL_FORMAT_FRACTAL_NZ)
layer.w2_weight.data = torch_npu.npu_format_cast(
layer.w2_weight.data, ACL_FORMAT_FRACTAL_NZ)
layer.w13_weight.data = self.pack_to_int32(layer.w13_weight.data)
layer.w2_weight.data = self.pack_to_int32(layer.w2_weight.data)

2
vllm_ascend/quantization/w8a8.py
View File

@@ -347,7 +347,7 @@ class AscendW8A8FusedMoEMethod:
# converting ACL_FORMAT_FRACTAL_NZ.
# npu_quant_grouped_matmul_dequant in eager mode does not accept
# ACL_FORMAT_FRACTAL_NZ.
if not is_310p() and is_enable_nz():
if not is_310p():
layer.w13_weight.data = torch_npu.npu_format_cast(
layer.w13_weight.data, ACL_FORMAT_FRACTAL_NZ).contiguous()
layer.w2_weight.data = torch_npu.npu_format_cast(

5
vllm_ascend/quantization/w8a8_dynamic.py
View File

@@ -270,9 +270,8 @@ class AscendW8A8DynamicFusedMoEMethod:
1, 2).contiguous()
layer.w2_weight.data = layer.w2_weight.data.transpose(
1, 2).contiguous()
if is_enable_nz():
torch_npu.npu_format_cast_(layer.w13_weight, ACL_FORMAT_FRACTAL_NZ)
torch_npu.npu_format_cast_(layer.w2_weight, ACL_FORMAT_FRACTAL_NZ)
torch_npu.npu_format_cast_(layer.w13_weight, ACL_FORMAT_FRACTAL_NZ)
torch_npu.npu_format_cast_(layer.w2_weight, ACL_FORMAT_FRACTAL_NZ)
layer.w13_weight_scale.data = layer.w13_weight_scale.data.view(
layer.w13_weight_scale.data.shape[0], -1)
layer.w13_weight_scale_fp32 = layer.w13_weight_scale.data.to(

7
vllm_ascend/utils.py
View File

@@ -81,10 +81,9 @@ def is_enable_nz(dtype: Optional[torch.dtype] = torch.int8,
"vllm_config must be provided when _ENABLE_NZ is None")
_ENABLE_NZ = envs_ascend.VLLM_ASCEND_ENABLE_NZ and vllm_config.model_config.hf_config.model_type != "qwen3_next"
_IS_EAGLE_MODE = (
vllm_config.speculative_config is not None and
getattr(vllm_config.speculative_config, 'method', None) in ("eagle", "eagle3")
)
_IS_EAGLE_MODE = (vllm_config.speculative_config is not None
and getattr(vllm_config.speculative_config, 'method',
None) in ("eagle", "eagle3"))
if dtype in [torch.float16, torch.bfloat16]:
return _ENABLE_NZ if _IS_EAGLE_MODE else False