Fix Transformers backend FP8 MoE and remove some boilerplate (#46820)

Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
This commit is contained in:
Harry Mellor
2026-06-27 00:16:05 +01:00
committed by GitHub
parent 2ff76a5e85
commit d8eb734d94
29 changed files with 80 additions and 156 deletions
+2 -3
View File
@@ -68,7 +68,7 @@ from vllm.model_executor.model_loader.weight_utils import (
from vllm.sequence import IntermediateTensors
from vllm.transformers_utils.configs.hy_v3 import HYV3Config
from .interfaces import SupportsLoRA, SupportsPP
from .interfaces import MixtureOfExperts, SupportsLoRA, SupportsPP
from .utils import (
AutoWeightsLoader,
PPMissingLayer,
@@ -392,7 +392,7 @@ class HYV3DecoderLayer(nn.Module):
@support_torch_compile
class HYV3Model(nn.Module):
class HYV3Model(nn.Module, MixtureOfExperts):
def __init__(self, *, vllm_config: VllmConfig, prefix: str = ""):
super().__init__()
@@ -429,7 +429,6 @@ class HYV3Model(nn.Module):
)
# Set MoE hyperparameters
self.expert_weights = []
self.num_expert_groups = 1
self.moe_layers = []
example_layer = None