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

tests/samplers/test_no_bad_words.py

 # SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 """Make sure bad_words works.
 
 Run `pytest tests/samplers/test_no_bad_words.py`.
@@ -103,7 +104,7 @@ class TestTwoTokenBadWord:
                                                 add_special_tokens=False)[0]
 
     def test_two_token_bad_word(self, vllm_runner):
-        with vllm_runner(self.MODEL) as llm:
+        with vllm_runner(self.MODEL, dtype="half") as llm:
             output_token_ids = self._generate(llm)
             assert output_token_ids[:2] == [
                 self.target_token_id1, self.target_token_id2

tests/samplers/test_ranks.py

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

tests/samplers/test_rejection_sampler.py

 # SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 """Tests for rejection sampling."""
 
 import pytest

tests/samplers/test_sampler.py

 # SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 
 import itertools
 import random

tests/samplers/test_seeded_generate.py

 # SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 """Verify that seeded random sampling is deterministic.
 
 Run `pytest tests/samplers/test_seeded_generate.py`.

tests/samplers/test_typical_acceptance_sampler.py

 # SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 """Tests for rejection sampling."""
 
 import pytest

tests/spec_decode/conftest.py

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

tests/spec_decode/e2e/conftest.py

 # SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 
 from collections.abc import Sequence
 from itertools import cycle

tests/spec_decode/e2e/test_compatibility.py

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

tests/spec_decode/e2e/test_eagle_correctness.py

 # SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 """This docstring details important information on the testing methodology.
 
 Most of the tests rely on "greedy equality", where we expect the output of

tests/spec_decode/e2e/test_integration.py

 # SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 """Tests which cover integration of the speculative decoding framework with
 other features, e.g. cuda graphs.
 """

tests/spec_decode/e2e/test_integration_dist_tp2.py

 # SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 """Tests which cover integration of the speculative decoding framework with
 tensor parallelism.
 """

tests/spec_decode/e2e/test_integration_dist_tp4.py

 # SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 """Tests which cover integration of the speculative decoding framework with
 tensor parallelism.
 """

tests/spec_decode/e2e/test_logprobs.py

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

tests/spec_decode/e2e/test_medusa_correctness.py

 # SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 """This docstring details important information on the testing methodology.
 
 Most of the tests rely on "greedy equality", where we expect the output of

tests/spec_decode/e2e/test_mlp_correctness.py

 # SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 """This docstring details important information on the testing methodology.
 
 Most of the tests rely on "greedy equality", where we expect the output of

tests/spec_decode/e2e/test_mtp_correctness.py

 # SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 """This docstring details important information on the testing methodology.
 
 Most of the tests rely on "greedy equality", where we expect the output of

tests/spec_decode/e2e/test_multistep_correctness.py

 # SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 """The tests in this file verify end-to-end speculative decoding correctness.
 
 This docstring details important information on the testing methodology.

tests/spec_decode/e2e/test_ngram_correctness.py

 # SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 """This docstring details important information on the testing methodology.
 
 Most of the tests rely on "greedy equality", where we expect the output of

tests/spec_decode/e2e/test_seed.py

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