[CI/Build] drop support for Python 3.8 EOL (#8464)

Signed-off-by: Aaron Pham <contact@aarnphm.xyz>

vllm/model_executor/models/mllama.py

-# coding=utf-8
 # Copyright 2024 the HuggingFace Inc. team. All rights reserved.
 #
 # Licensed under the Apache License, Version 2.0 (the "License");

vllm/model_executor/models/mlp_speculator.py

@@ -37,7 +37,7 @@ class MLPSpeculatorLayerNorm(nn.Module):
         eps=1e-06,
         elementwise_scale_and_shift=True,
     ):
-        super(MLPSpeculatorLayerNorm, self).__init__()
+        super().__init__()
         self.elementwise_scale_and_shift = elementwise_scale_and_shift
         if self.elementwise_scale_and_shift:
             self.weight = nn.Parameter(torch.empty(normalized_shape))

vllm/model_executor/models/molmo.py

@@ -1121,9 +1121,9 @@ class MolmoForCausalLM(nn.Module, SupportsMultiModal, SupportsPP):
             batch_size * num_image * num_patch, -1).contiguous()
 
         image_input_idx = image_input_idx * valid.to(image_input_idx.dtype)
-        offset = torch.cat(
-            [seq_len.new_zeros(
-                (1)), seq_len.cumsum(dim=0)[:-1]], dim=0)[:, None]
+        offset = torch.cat([seq_len.new_zeros(1),
+                            seq_len.cumsum(dim=0)[:-1]],
+                           dim=0)[:, None]
         image_input_idx = image_input_idx + offset.to(image_input_idx.dtype)
         image_input_idx = image_input_idx.flatten()[:, None]
         mat = image_input_idx == torch.arange(

vllm/model_executor/models/mpt.py

-# coding=utf-8
 # Adapted from https://huggingface.co/mosaicml/mpt-7b/tree/main
 import math
 from typing import Iterable, List, Optional, Tuple, Union

vllm/model_executor/models/nemotron.py

-# coding=utf-8
 # Adapted from
 # https://github.com/huggingface/transformers/blob/v4.28.0/src/transformers/models/llama/modeling_llama.py
 # Copyright 2023 The vLLM team.

vllm/model_executor/models/olmo.py

-# coding=utf-8
 # Adapted from
 # https://github.com/huggingface/transformers/blob/v4.40.1/src/transformers/models/olmo/modeling_olmo.py
 # Copyright 2024 The vLLM team.

vllm/model_executor/models/opt.py

-# coding=utf-8
 # Adapted from
 # https://github.com/huggingface/transformers/blob/v4.28.0/src/transformers/models/opt/modeling_opt.py
 # Copyright 2023 The vLLM team.

vllm/model_executor/models/orion.py

-# coding=utf-8
 # Adapted from
 # https://huggingface.co/OrionStarAI/Orion-14B-Base/blob/main/modeling_orion.py
 # Copyright (c) OrionStar Inc.

vllm/model_executor/models/persimmon.py

-# coding=utf-8
 # adapted from https://github.com/huggingface/transformers/blob/v4.39.3/src/transformers/models/persimmon/modeling_persimmon.py
 # Copyright 2023 The vLLM team.
 # Copyright 2023 EleutherAI and the HuggingFace Inc. team. All rights reserved.

vllm/model_executor/models/phi.py

-# coding=utf-8
 # Adapted from
 # https://huggingface.co/microsoft/phi-1_5/blob/main/modeling_phi.py
 # Copyright 2023 The vLLM team.

vllm/model_executor/models/phi3.py

-# coding=utf-8
 # Adapted from llama.py
 """Inference-only Phi3 model code inherit from Llama.py"""
 

vllm/model_executor/models/phi3v.py

-# coding=utf-8
 # Copyright 2024 The vLLM team.
 # Copyright 2024 Microsoft and the HuggingFace Inc. team. All rights reserved.
 #

vllm/model_executor/models/phimoe.py

-# coding=utf-8
 # Adapted from
 # https://github.com/huggingface/transformers/blob/v4.28.0/src/transformers/models/llama/modeling_llama.py
 # Copyright 2023 The vLLM team.

vllm/model_executor/models/pixtral.py

@@ -136,11 +136,11 @@ def input_processor_for_pixtral(ctx: InputContext, inputs: DecoderOnlyInputs):
 
         if image_token_id not in inputs['prompt_token_ids']:
             raise ValueError(
-                (f"You've passed {inputs=} without {image_token_id=}"
+                f"You've passed {inputs=} without {image_token_id=}"
                 " Make sure to process your input via mistral_common's"
                 " tokenizer or pass a chat completion request. For more"
                 " For more info, see: "
-                 "https://github.com/vllm-project/vllm/issues/8411."))
+                "https://github.com/vllm-project/vllm/issues/8411.")
 
     return inputs
 

vllm/model_executor/models/qwen.py

-# coding=utf-8
 # Adapted from
 # https://huggingface.co/Qwen/Qwen-7B/blob/main/modeling_qwen.py
 # Copyright (c) Alibaba Cloud.

vllm/model_executor/models/qwen2.py

-# coding=utf-8
 # Adapted from
 # https://github.com/huggingface/transformers/blob/v4.28.0/src/transformers/models/qwen2/modeling_qwen2.py
 # Copyright 2024 The Qwen team.
@@ -417,9 +416,9 @@ class Qwen2ForCausalLM(nn.Module, SupportsLoRA, SupportsPP):
                 and hasattr(config, "max_window_layers")):
             raise ValueError("Sliding window for some but all layers is not "
                              "supported. This model uses sliding window "
-                             "but `max_window_layers` = %s is less than "
-                             "`num_hidden_layers` = %s. Please open an issue "
-                             "to discuss this feature." % (
+                             "but `max_window_layers` = {} is less than "
+                             "`num_hidden_layers` = {}. Please open an issue "
+                             "to discuss this feature.".format(
                                  config.max_window_layers,
                                  config.num_hidden_layers,
                              ))

vllm/model_executor/models/qwen2_audio.py

-# coding=utf-8
 # Copyright 2024 The Qwen team.
 # Copyright 2023 The vLLM team.
 # Copyright 2022 EleutherAI and the HuggingFace Inc. team. All rights reserved.

vllm/model_executor/models/qwen2_cls.py

-# coding=utf-8
 # Adapted from
 # https://huggingface.co/Qwen/Qwen2.5-Math-RM-72B/blob/main/modeling_qwen2_rm.py
 # Copyright 2024 Kakao Corp. (Kanana-X Team)
@@ -60,9 +59,9 @@ class Qwen2ForSequenceClassification(nn.Module):
                 and hasattr(config, "max_window_layers")):
             raise ValueError("Sliding window for some but all layers is not "
                              "supported. This model uses sliding window "
-                             "but `max_window_layers` = %s is less than "
-                             "`num_hidden_layers` = %s. Please open an issue "
-                             "to discuss this feature." % (
+                             "but `max_window_layers` = {} is less than "
+                             "`num_hidden_layers` = {}. Please open an issue "
+                             "to discuss this feature.".format(
                                  config.max_window_layers,
                                  config.num_hidden_layers,
                              ))

vllm/model_executor/models/qwen2_moe.py

-# coding=utf-8
 # Adapted from
 # https://github.com/huggingface/transformers/blob/v4.28.0/src/transformers/models/qwen2_moe/modeling_qwen2_moe.py
 # Copyright 2024 The Qwen team.

vllm/model_executor/models/qwen2_rm.py

-# coding=utf-8
 # Adapted from
 # https://huggingface.co/Qwen/Qwen2.5-Math-RM-72B/blob/main/modeling_qwen2_rm.py
 # Copyright 2024 The Qwen team.
@@ -71,9 +70,9 @@ class Qwen2ForRewardModel(nn.Module, SupportsPP):
                 and hasattr(config, "max_window_layers")):
             raise ValueError("Sliding window for some but all layers is not "
                              "supported. This model uses sliding window "
-                             "but `max_window_layers` = %s is less than "
-                             "`num_hidden_layers` = %s. Please open an issue "
-                             "to discuss this feature." % (
+                             "but `max_window_layers` = {} is less than "
+                             "`num_hidden_layers` = {}. Please open an issue "
+                             "to discuss this feature.".format(
                                  config.max_window_layers,
                                  config.num_hidden_layers,
                              ))