Merge tag 'v0.9.1' into v0.9.1-ori

examples/offline_inference/basic/embed.py

 # SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 
 from argparse import Namespace
 

examples/offline_inference/basic/generate.py

 # SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 
 from vllm import LLM, EngineArgs
 from vllm.utils import FlexibleArgumentParser

examples/offline_inference/basic/score.py

 # SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 
 from argparse import Namespace
 

examples/offline_inference/batch_llm_inference.py

 # SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 """
 This example shows how to use Ray Data for data parallel batch inference.
 

examples/offline_inference/chat_with_tools.py

 # SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 
 # ruff: noqa
 import json

examples/offline_inference/context_extension.py

+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+"""
+This script demonstrates how to extend the context length
+of a Qwen model using the YARN method (rope_scaling)
+and run a simple chat example.
+
+Usage:
+    python examples/offline_inference/context_extension.py
+"""
+
+from vllm import LLM, SamplingParams
+
+
+def create_llm():
+    rope_theta = 1000000
+    original_max_position_embeddings = 32768
+    factor = 4.0
+
+    # Use yarn to extend context
+    hf_overrides = {
+        "rope_theta": rope_theta,
+        "rope_scaling": {
+            "rope_type": "yarn",
+            "factor": factor,
+            "original_max_position_embeddings": original_max_position_embeddings,
+        },
+        "max_model_len": int(original_max_position_embeddings * factor),
+    }
+
+    llm = LLM(model="Qwen/Qwen3-0.6B", hf_overrides=hf_overrides)
+    return llm
+
+
+def run_llm_chat(llm):
+    sampling_params = SamplingParams(
+        temperature=0.8,
+        top_p=0.95,
+        max_tokens=128,
+    )
+
+    conversation = [
+        {"role": "system", "content": "You are a helpful assistant"},
+        {"role": "user", "content": "Hello"},
+        {"role": "assistant", "content": "Hello! How can I assist you today?"},
+    ]
+    outputs = llm.chat(conversation, sampling_params, use_tqdm=False)
+    return outputs
+
+
+def print_outputs(outputs):
+    print("\nGenerated Outputs:\n" + "-" * 80)
+    for output in outputs:
+        prompt = output.prompt
+        generated_text = output.outputs[0].text
+        print(f"Prompt: {prompt!r}\n")
+        print(f"Generated text: {generated_text!r}")
+        print("-" * 80)
+
+
+def main():
+    llm = create_llm()
+    outputs = run_llm_chat(llm)
+    print_outputs(outputs)
+
+
+if __name__ == "__main__":
+    main()

examples/offline_inference/data_parallel.py

 # SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 """
 Usage:
 Single node:
@@ -97,10 +98,14 @@ def main(
     # with DP, each rank should process different prompts.
     # usually all the DP ranks process a full dataset,
     # and each rank processes a different part of the dataset.
-    promts_per_rank = len(prompts) // dp_size
-    start = global_dp_rank * promts_per_rank
-    end = start + promts_per_rank
-    prompts = prompts[start:end]
+    floor = len(prompts) // dp_size
+    remainder = len(prompts) % dp_size
+
+    # Distribute prompts into even groups.
+    def start(rank):
+        return rank * floor + min(rank, remainder)
+
+    prompts = prompts[start(global_dp_rank) : start(global_dp_rank + 1)]
     if len(prompts) == 0:
         # if any rank has no prompts to process,
         # we need to set a placeholder prompt

examples/offline_inference/disaggregated-prefill-v1/decode_example.py

 # SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 
 from vllm import LLM, SamplingParams
 from vllm.config import KVTransferConfig

examples/offline_inference/disaggregated-prefill-v1/prefill_example.py

 # SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 
 from vllm import LLM, SamplingParams
 from vllm.config import KVTransferConfig

examples/offline_inference/disaggregated-prefill-v1/run.sh

 rm -rf local_storage/
-rm output.txt
 
-VLLM_ENABLE_V1_MULTIPROCESSING=0 CUDA_VISIBLE_DEVICES=0 python3 prefill_example.py
-VLLM_ENABLE_V1_MULTIPROCESSING=0 CUDA_VISIBLE_DEVICES=0 python3 decode_example.py
+if [ -f "output.txt" ]; then
+    rm output.txt
+fi
+
+# The directory of current script
+SCRIPT_DIR=$(dirname "$(readlink -f "$0")")
+
+VLLM_ENABLE_V1_MULTIPROCESSING=0 CUDA_VISIBLE_DEVICES=0 python3 "$SCRIPT_DIR/prefill_example.py"
+VLLM_ENABLE_V1_MULTIPROCESSING=0 CUDA_VISIBLE_DEVICES=0 python3 "$SCRIPT_DIR/decode_example.py"

examples/offline_inference/disaggregated_prefill.py

 # SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 """
 This file demonstrates the example usage of disaggregated prefilling
 We will launch 2 vllm instances (GPU 0 for prefill and GPU 1 for decode),

examples/offline_inference/eagle.py

 # SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 import argparse
 import json
 import os

examples/offline_inference/embed_jina_embeddings_v3.py

 # SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 
 from argparse import Namespace
 

examples/offline_inference/embed_matryoshka_fy.py

 # SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 
 from argparse import Namespace
 

examples/offline_inference/encoder_decoder.py

 # SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 """
 Demonstrate prompting of text-to-text
 encoder/decoder models, specifically BART

examples/offline_inference/encoder_decoder_multimodal.py

 # SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 """
 This example shows how to use vLLM for running offline inference with
 the explicit/implicit prompt format on enc-dec LMMs for text generation.

examples/offline_inference/llm_engine_example.py

 # SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 """
 This file demonstrates using the `LLMEngine`
 for processing prompts with various sampling parameters.

examples/offline_inference/load_sharded_state.py

 # SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 """
 Validates the loading of a model saved with the sharded_state format.
 This script demonstrates how to load a model that was previously saved

examples/offline_inference/lora_with_quantization_inference.py

 # SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 """
 This example shows how to use LoRA with different quantization techniques
 for offline inference.

examples/offline_inference/metrics.py

 # SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 
 from vllm import LLM, SamplingParams
 from vllm.v1.metrics.reader import Counter, Gauge, Histogram, Vector