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

Update deprecated type hinting in models (#18132)

Browse Source 
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
This commit is contained in:
Harry Mellor
2025-05-15 06:06:50 +01:00
committed by GitHub
parent 83f74c698f
commit 26d0419309
130 changed files with 971 additions and 901 deletions
Download Patch File Download Diff File Expand all files Collapse all files

17
vllm/model_executor/models/qwen2_moe.py
View File

@@ -23,7 +23,8 @@
# See the License for the specific language governing permissions and
# limitations under the License.
"""Inference-only Qwen2MoE model compatible with HuggingFace weights."""
from typing import Any, Dict, Iterable, Optional, Set, Tuple, Union
from collections.abc import Iterable
from typing import Any, Optional, Union
import torch
import torch.nn.functional as F
@@ -169,12 +170,12 @@ class Qwen2MoeAttention(nn.Module):
num_heads: int,
num_kv_heads: int,
rope_theta: float = 10000,
rope_scaling: Optional[Dict[str, Any]] = None,
rope_scaling: Optional[dict[str, Any]] = None,
max_position_embeddings: int = 8192,
cache_config: Optional[CacheConfig] = None,
quant_config: Optional[QuantizationConfig] = None,
prefix: str = "",
dual_chunk_attention_config: Optional[Dict[str, Any]] = None,
dual_chunk_attention_config: Optional[dict[str, Any]] = None,
) -> None:
super().__init__()
self.hidden_size = hidden_size
@@ -389,8 +390,8 @@ class Qwen2MoeModel(nn.Module):
hidden_states, _ = self.norm(hidden_states, residual)
return hidden_states
def load_weights(self, weights: Iterable[Tuple[str,
torch.Tensor]]) -> Set[str]:
def load_weights(self, weights: Iterable[tuple[str,
torch.Tensor]]) -> set[str]:
stacked_params_mapping = [
# (param_name, shard_name, shard_id)
("qkv_proj", "q_proj", "q"),
@@ -409,7 +410,7 @@ class Qwen2MoeModel(nn.Module):
num_experts=self.config.num_experts)
params_dict = dict(self.named_parameters())
loaded_params: Set[str] = set()
loaded_params: set[str] = set()
for name, loaded_weight in weights:
for (param_name, weight_name, shard_id) in stacked_params_mapping:
# Skip non-stacked layers and experts (experts handled below).
@@ -532,8 +533,8 @@ class Qwen2MoeForCausalLM(nn.Module, SupportsPP):
sampling_metadata)
return logits
def load_weights(self, weights: Iterable[Tuple[str,
torch.Tensor]]) -> Set[str]:
def load_weights(self, weights: Iterable[tuple[str,
torch.Tensor]]) -> set[str]:
loader = AutoWeightsLoader(
self,
skip_prefixes=(["rotary_emb.inv_freq"]),