chore (1.x): support vllm 0.18.0 without backwards compat

.github/workflows/test.yml

@@ -84,7 +84,7 @@ jobs:
           # Lower bound support
           - vllm_version:
               name: "vLLM:lowest"
-              repo: "git+https://github.com/vllm-project/vllm --tag v0.11.0"
+              repo: "git+https://github.com/vllm-project/vllm --tag v0.17.0"
             test_suite:
               name: "backward compat"
               markers: "compat or (cpu and basic and not quantized and not sb)"
@@ -94,86 +94,9 @@ jobs:
             os: "ubuntu-latest"
             python_version: "3.12"
           # Intermediate versions of vllm to check basic support for as well
-          - vllm_version:
-              name: "vLLM:0.11.1"
-              repo: "git+https://github.com/vllm-project/vllm --tag v0.11.1"
-            test_suite:
-              name: "backward compat"
-              markers: "cpu and basic and not quantized and not sb"
-              flags: "--timeout=300"
-              hf_model_2: "sentence-transformers/all-roberta-large-v1"
-              hf_model_2_rev: "cf74d8acd4f198de950bf004b262e6accfed5d2c"
-            os: "ubuntu-latest"
-            python_version: "3.12"
-          - vllm_version:
-              name: "vLLM:0.11.2"
-              repo: "git+https://github.com/vllm-project/vllm --tag v0.11.2"
-            test_suite:
-              name: "backward compat"
-              markers: "cpu and basic and not quantized and not sb"
-              flags: "--timeout=300"
-              hf_model_2: "sentence-transformers/all-roberta-large-v1"
-              hf_model_2_rev: "cf74d8acd4f198de950bf004b262e6accfed5d2c"
-            os: "ubuntu-latest"
-            python_version: "3.12"
-          - vllm_version:
-              name: "vLLM:0.12.0"
-              repo: "git+https://github.com/vllm-project/vllm --tag v0.12.0"
-            test_suite:
-              name: "backward compat"
-              markers: "cpu and basic and not quantized and not sb"
-              flags: "--timeout=300"
-              hf_model_2: "sentence-transformers/all-roberta-large-v1"
-              hf_model_2_rev: "cf74d8acd4f198de950bf004b262e6accfed5d2c"
-            os: "ubuntu-latest"
-            python_version: "3.12"
-          - vllm_version:
-              name: "vLLM:0.13.0"
-              repo: "git+https://github.com/vllm-project/vllm --tag v0.13.0"
-            test_suite:
-              name: "backward compat"
-              markers: "cpu and basic and not quantized and not sb"
-              flags: "--timeout=300"
-              hf_model_2: "sentence-transformers/all-roberta-large-v1"
-              hf_model_2_rev: "cf74d8acd4f198de950bf004b262e6accfed5d2c"
-            os: "ubuntu-latest"
-            python_version: "3.12"
-          - vllm_version:
-              name: "vLLM:0.14.0"
-              repo: "git+https://github.com/vllm-project/vllm --tag v0.14.0"
-            test_suite:
-              name: "backward compat"
-              markers: "cpu and basic and not quantized and not sb"
-              flags: "--timeout=300"
-              hf_model_2: "sentence-transformers/all-roberta-large-v1"
-              hf_model_2_rev: "cf74d8acd4f198de950bf004b262e6accfed5d2c"
-            os: "ubuntu-latest"
-            python_version: "3.12"
-          - vllm_version:
-              name: "vLLM:0.14.1"
-              repo: "git+https://github.com/vllm-project/vllm --tag v0.14.1"
-            test_suite:
-              name: "backward compat"
-              markers: "cpu and basic and not quantized and not sb"
-              flags: "--timeout=300"
-              hf_model_2: "sentence-transformers/all-roberta-large-v1"
-              hf_model_2_rev: "cf74d8acd4f198de950bf004b262e6accfed5d2c"
-            os: "ubuntu-latest"
-            python_version: "3.12"
-          - vllm_version:
-              name: "vLLM:0.15.0"
-              repo: "git+https://github.com/vllm-project/vllm --tag v0.15.0"
-            test_suite:
-              name: "backward compat"
-              markers: "cpu and basic and not quantized and not sb"
-              flags: "--timeout=300"
-              hf_model_2: "sentence-transformers/all-roberta-large-v1"
-              hf_model_2_rev: "cf74d8acd4f198de950bf004b262e6accfed5d2c"
-            os: "ubuntu-latest"
-            python_version: "3.12"
           - vllm_version:
               name: "vLLM:0.15.1"
-              repo: "git+https://github.com/vllm-project/vllm --tag v0.15.1"
+              repo: "git+https://github.com/vllm-project/vllm --tag v0.17.1"
             test_suite:
               name: "backward compat"
               markers: "cpu and basic and not quantized and not sb"

pyproject.toml

@@ -13,7 +13,7 @@ license = {text = "Apache 2"}
 dependencies = [
     "fms-model-optimizer[fp8]>=0.8.0",
     "ibm-fms>=1.7.0,<2.0",
-    "vllm>=0.11.0,<0.16.1",
+    "vllm>=0.17.0,<0.18.1",
 ]
 requires-python = ">=3.11"
 dynamic = ["version"]
@@ -70,7 +70,7 @@ environments = [
 ]
 
 [tool.uv.sources]
-vllm = { git = "https://github.com/vllm-project/vllm", rev = "v0.16.0" }
+vllm = { git = "https://github.com/vllm-project/vllm", rev = "v0.18.0" }
 
 [tool.ty.rules]
 possibly-missing-attribute = "ignore"

tests/conftest.py

@@ -214,6 +214,8 @@ def cleanup_fixture(should_do_global_cleanup_after_test: bool):
     SpyrePlatform._used_with_cli = False
     yield
     if should_do_global_cleanup_after_test:
+        # Workaround torch.accelerator.empty_cache for torch 2.7.1 and vllm v0.18.0 compatibility
+        setattr(torch.accelerator, "empty_cache", lambda: None)  # noqa
         cleanup_dist_env_and_memory()
 
 

tests/e2e/test_chunked_prefill_tkv_steps.py

@@ -119,9 +119,6 @@ def make_scheduler_output(
         scheduled_cached_reqs=scheduled_cached_reqs,
         num_scheduled_tokens=num_scheduled_tokens,
         total_num_scheduled_tokens=total_tokens,
-        scheduled_spec_decode_tokens={},
-        scheduled_encoder_inputs={},
-        num_common_prefix_blocks=[],
         finished_req_ids=finished_req_ids,
         kv_connector_metadata=None,
         **extra_args,
@@ -134,7 +131,6 @@ def make_new_request_data(req_id, prompt_len):
         prompt_token_ids=[42] * prompt_len,
         sampling_params=SamplingParams(),
         pooling_params=None,
-        eos_token_id=None,
     )
     return NewRequestData.from_request(req, block_ids=[])
 

tests/scheduling_utils.py

@@ -120,7 +120,6 @@ def create_request_for_scheduler_test(
         request_id=str(request_id),
         sampling_params=sampling_params,
         prompt_token_ids=prompt,
-        eos_token_id=None,
         arrival_time=0,
         lora_request=None,
         pooling_params=None,

tests/spyre_util.py

@@ -22,13 +22,8 @@
 from vllm_spyre.platform import SpyrePlatform
 from vllm_spyre import envs
 
-try:
-    # old
-    from vllm.utils import FlexibleArgumentParser, get_open_port
-except ImportError:
-    # new
-    from vllm.utils.argparse_utils import FlexibleArgumentParser
-    from vllm.utils.network_utils import get_open_port
+from vllm.utils.argparse_utils import FlexibleArgumentParser
+from vllm.utils.network_utils import get_open_port
 
 from vllm.v1.request import Request
 
@@ -448,7 +443,6 @@ def create_random_request(
         request_id=str(request_id),
         prompt_token_ids=prompt_token_ids,
         sampling_params=sampling_params,
-        eos_token_id=None,
         arrival_time=0,
         lora_request=None,
         pooling_params=None,

tests/utils/test_cli_args.py

@@ -10,12 +10,7 @@
 from vllm_spyre.config.model_registry import get_model_registry
 from spyre_util import environ_checkpoint, REFERENCE_MODELS
 
-try:
-    # old
-    from vllm.utils import FlexibleArgumentParser
-except ImportError:
-    # new
-    from vllm.utils.argparse_utils import FlexibleArgumentParser
+from vllm.utils.argparse_utils import FlexibleArgumentParser
 
 global_default = 192
 
@@ -69,8 +64,6 @@ def sendnn_configured() -> bool:
         "32",
         "-tp",
         "4",
-        "--swap-space",  # to prevent a validation error in the 16GB memory test env.
-        "1",
     ]
 
     if model_name == "ibm-granite/granite-3.3-8b-instruct":

tests/utils/test_platform_validation.py

@@ -4,15 +4,29 @@
 from SamplingParams during request validation.
 """
 
+from unittest.mock import MagicMock
 import pytest
+
 from vllm import SamplingParams
+from vllm.inputs.data import token_inputs
 from vllm.pooling_params import PoolingParams
 from vllm.sampling_params import StructuredOutputsParams
 from vllm_spyre.platform import SpyrePlatform
 
 pytestmark = pytest.mark.skip_global_cleanup
 
 
+@pytest.fixture(autouse=True)
+def mock_spyre_config():
+    """Mock SpyrePlatform._config for all tests."""
+    original_config = SpyrePlatform._config
+    mock_config = MagicMock()
+    mock_config.model_config.max_model_len = 512
+    SpyrePlatform._config = mock_config
+    yield mock_config
+    SpyrePlatform._config = original_config
+
+
 class TestStructuredOutputValidation:
     """Test that platform validation strips structured outputs from requests."""
 
@@ -24,7 +38,8 @@ def test_strips_structured_outputs(self):
 
         assert params.structured_outputs is not None
 
-        SpyrePlatform.validate_request("Test prompt", params)
+        processed_inputs = token_inputs(prompt_token_ids=[1, 2, 3])
+        SpyrePlatform.validate_request(processed_inputs, params)
 
         assert params.structured_outputs is None
 
@@ -34,7 +49,8 @@ def test_logs_warning_when_stripping(self, caplog_vllm_spyre):
             max_tokens=20, structured_outputs=StructuredOutputsParams(json_object=True)
         )
 
-        SpyrePlatform.validate_request("Test prompt", params)
+        processed_inputs = token_inputs(prompt_token_ids=[1, 2, 3])
+        SpyrePlatform.validate_request(processed_inputs, params)
 
         assert len(caplog_vllm_spyre.records) > 0
         warning_record = caplog_vllm_spyre.records[0]
@@ -55,7 +71,8 @@ def test_strips_different_structured_output_types(self, structured_output):
 
         assert params.structured_outputs is not None
 
-        SpyrePlatform.validate_request("Test prompt", params)
+        processed_inputs = token_inputs(prompt_token_ids=[1, 2, 3])
+        SpyrePlatform.validate_request(processed_inputs, params)
 
         assert params.structured_outputs is None
 
@@ -77,7 +94,8 @@ def test_preserves_other_sampling_params(self):
             "top_k": params.top_k,
         }
 
-        SpyrePlatform.validate_request("Test prompt", params)
+        processed_inputs = token_inputs(prompt_token_ids=[1, 2, 3])
+        SpyrePlatform.validate_request(processed_inputs, params)
 
         # Verify other params are unchanged
         assert params.max_tokens == original_values["max_tokens"]
@@ -92,7 +110,8 @@ def test_does_not_affect_pooling_params(self):
         pooling_params = PoolingParams()
 
         # Should not raise any errors and should return early
-        SpyrePlatform.validate_request("Test prompt", pooling_params)
+        processed_inputs = token_inputs(prompt_token_ids=[1, 2, 3])
+        SpyrePlatform.validate_request(processed_inputs, pooling_params)
 
         # PoolingParams don't have structured_outputs, so just verify no exception
         assert True  # If we got here, the early return worked