[Quantization] Auto-disable DeepGEMM for Kolibri 1 on SM12x (#61274)

Signed-off-by: Luca Motz <luca.motz@icloud.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
Luca Motz
2026-10-14 12:59:50 +00:00
committed by GitHub
co-authored by Claude Opus 5.5
parent 3723a1ea96
commit 2d09fdcdc2
3 changed files with 22 additions and 13 deletions
@@ -9,6 +9,7 @@ import torch
import vllm.model_executor.layers.fused_moe.modular_kernel as mk
from vllm import envs
from vllm._aiter_ops import rocm_aiter_ops
from vllm.config import get_current_vllm_config
from vllm.config.kernel import MoEBackend
from vllm.logger import init_logger
from vllm.model_executor.layers.fused_moe.all2all_utils import (
@@ -40,6 +41,7 @@ from vllm.model_executor.layers.quantization.utils.quant_utils import (
)
from vllm.model_executor.utils import replace_parameter
from vllm.platforms import current_platform
from vllm.utils.deep_gemm import should_auto_disable_deep_gemm
from vllm.utils.math_utils import round_up
logger = init_logger(__name__)
@@ -492,6 +494,13 @@ def select_fp8_moe_backend(
return _return_or_raise(
backend, config, weight_key, activation_key, activation_format
)
elif (model_config := get_current_vllm_config().model_config) is not None and (
should_auto_disable_deep_gemm(
getattr(model_config.hf_text_config, "model_type", None)
)
):
AVAILABLE_BACKENDS.remove(Fp8MoeBackend.DEEPGEMM)
AVAILABLE_BACKENDS.remove(Fp8MoeBackend.BATCHED_DEEPGEMM)
# Handle explicit AITER FP8 configuration.
if envs.is_set("VLLM_ROCM_USE_AITER") or envs.is_set("VLLM_ROCM_USE_AITER_MOE"):