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

13
vllm/model_executor/models/jamba.py
View File

@@ -1,6 +1,7 @@
# SPDX-License-Identifier: Apache-2.0
"""Inference-only Jamba model."""
from typing import Iterable, Optional, Set, Tuple
from collections.abc import Iterable
from typing import Optional
import torch
from torch import nn
@@ -442,7 +443,7 @@ class JambaForCausalLM(nn.Module, HasInnerState, SupportsLoRA, SupportsPP,
return self.mamba_cache.get_seqlen_agnostic_capture_inputs(batch_size)
def _get_mamba_cache_shape(
self) -> Tuple[Tuple[int, int], Tuple[int, int]]:
self) -> tuple[tuple[int, int], tuple[int, int]]:
world_size = get_tensor_model_parallel_world_size()
hidden_size = self.config.hidden_size
conv_state_shape = (
@@ -464,8 +465,8 @@ class JambaForCausalLM(nn.Module, HasInnerState, SupportsLoRA, 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]:
stacked_params_mapping = [
# (param_name, shard_name, shard_id)
("qkv_proj", "q_proj", "q"),
@@ -482,7 +483,7 @@ class JambaForCausalLM(nn.Module, HasInnerState, SupportsLoRA, SupportsPP,
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:
if "rotary_emb.inv_freq" in name:
continue
@@ -583,7 +584,7 @@ class JambaForSequenceClassification(JambaForCausalLM):
logits = self.score(hidden_states)
return self._pooler(logits, pooling_metadata)
def load_weights(self, weights: Iterable[Tuple[str, torch.Tensor]]):
def load_weights(self, weights: Iterable[tuple[str, torch.Tensor]]):
# TODO: The reward weights themselves have float32 accuracy data, we
# would like to load them in fp32 to get that extra precision.
super().load_weights(weights)