Fix guide decode

Signed-off-by: Yikun Jiang <yikunkero@gmail.com>

Author: Yikun

.github/workflows/vllm_ascend_test_full.yaml

@@ -134,7 +134,7 @@ jobs:
           pytest -sv tests/e2e/singlecard/test_camem.py
           pytest -sv tests/e2e/singlecard/test_chunked.py
           pytest -sv tests/e2e/singlecard/test_embedding.py
-          #pytest -sv tests/e2e/singlecard/test_guided_decoding.py
+          pytest -sv tests/e2e/singlecard/test_guided_decoding.py
           #pytest -sv tests/e2e/singlecard/test_ilama_lora.py
           pytest -sv tests/e2e/singlecard/test_profile_execute_duration.py
           pytest -sv tests/e2e/singlecard/test_quantization.py

tests/e2e/singlecard/test_guided_decoding.py

@@ -18,21 +18,19 @@
 #
 import json
 import os
-from typing import Any, Dict
 
 import jsonschema
 import pytest
 import regex as re
-from vllm.outputs import RequestOutput
-from vllm.sampling_params import SamplingParams
 
 from vllm_ascend.utils import vllm_version_is
 
 if vllm_version_is("0.10.2"):
-    from vllm.sampling_params import \
-        GuidedDecodingParams as StructuredOutputsParams
+    from vllm.sampling_params import GuidedDecodingParams, SamplingParams
 else:
-    from vllm.sampling_params import StructuredOutputsParams
+    from vllm.sampling_params import SamplingParams, StructuredOutputsParams
+
+from vllm.outputs import RequestOutput
 
 from tests.e2e.conftest import VllmRunner
 
@@ -90,36 +88,31 @@ def sample_json_schema():
     }
 
 
-def construct_sampling_params(
-        struct_param, sampling_kwargs: Dict[str, Any]) -> SamplingParams:
-    if vllm_version_is("0.10.2"):
-        return SamplingParams(guided_decoding=struct_param, **sampling_kwargs)
-    else:
-        return SamplingParams(structured_outputs=struct_param,
-                              **sampling_kwargs)
-
-
 @pytest.mark.parametrize("guided_decoding_backend", GuidedDecodingBackend)
 def test_guided_json_completion(guided_decoding_backend: str,
                                 sample_json_schema):
-    struct_output_params = StructuredOutputsParams(json=sample_json_schema, )
-    sampling_params = construct_sampling_params(struct_output_params, {
-        "temperature": 1.0,
-        "max_tokens": 500,
-    })
-
-    runner_kwargs: Dict[str, Any] = {
-        "seed": 0,
-    }
     if vllm_version_is("0.10.2"):
-        runner_kwargs["guided_decoding_backend"] = guided_decoding_backend
+        sampling_params = SamplingParams(
+            temperature=1.0,
+            max_tokens=500,
+            guided_decoding=GuidedDecodingParams(json=sample_json_schema))
+        runner_kwargs = {
+            "seed": 0,
+            "guided_decoding_backend": guided_decoding_backend,
+        }
     else:
-        runner_kwargs["structured_outputs_config"] = {
-            "backend": guided_decoding_backend
+        sampling_params = SamplingParams(
+            temperature=1.0,
+            max_tokens=500,
+            structured_outputs=StructuredOutputsParams(
+                json=sample_json_schema))
+        runner_kwargs = {
+            "seed": 0,
+            "structured_outputs_config": {
+                "backend": guided_decoding_backend
+            },
         }
-
-    with VllmRunner(MODEL_NAME,
-                    **runner_kwargs) as vllm_model:  # type: ignore[arg-type]
+    with VllmRunner(MODEL_NAME, **runner_kwargs) as vllm_model:
         prompts = [
             f"Give an example JSON for an employee profile "
             f"that fits this schema: {sample_json_schema}"
@@ -147,21 +140,25 @@ def test_guided_json_completion(guided_decoding_backend: str,
 def test_guided_regex(guided_decoding_backend: str, sample_regex):
     if guided_decoding_backend == "outlines":
         pytest.skip("Outlines doesn't support regex-based guided decoding.")
-
-    struct_output_params = StructuredOutputsParams(json=sample_regex, )
-    sampling_params = construct_sampling_params(struct_output_params, {
-        "temperature": 0.8,
-        "top_p": 0.95,
-    })
-
-    runner_kwargs: Dict[str, Any] = {
-        "seed": 0,
-    }
     if vllm_version_is("0.10.2"):
-        runner_kwargs["guided_decoding_backend"] = guided_decoding_backend
+        sampling_params = SamplingParams(
+            temperature=0.8,
+            top_p=0.95,
+            guided_decoding=GuidedDecodingParams(regex=sample_regex))
+        runner_kwargs = {
+            "seed": 0,
+            "guided_decoding_backend": guided_decoding_backend,
+        }
     else:
-        runner_kwargs["structured_outputs_config"] = {
-            "backend": guided_decoding_backend
+        sampling_params = SamplingParams(
+            temperature=0.8,
+            top_p=0.95,
+            structured_outputs=StructuredOutputsParams(regex=sample_regex))
+        runner_kwargs = {
+            "seed": 0,
+            "structured_outputs_config": {
+                "backend": guided_decoding_backend
+            },
         }
 
     with VllmRunner(MODEL_NAME, **runner_kwargs) as vllm_model: