[Quantization] Deprecate Long Tail of Schemes (#31688)

Signed-off-by: Robert Shaw <robshaw@redhat.com>
Signed-off-by: Robert Shaw <114415538+robertgshaw2-redhat@users.noreply.github.com>
Co-authored-by: Robert Shaw <robshaw@redhat.com>
Co-authored-by: Wentao Ye <44945378+yewentao256@users.noreply.github.com>

tests/quantization/test_auto_round.py

@@ -26,7 +26,9 @@ MODELS = [
 )
 @pytest.mark.parametrize("model", MODELS)
 def test_auto_round(vllm_runner, model):
-    with vllm_runner(model, enforce_eager=True) as llm:
+    with vllm_runner(
+        model, enforce_eager=True, allow_deprecated_quantization=True
+    ) as llm:
         output = llm.generate_greedy(["The capital of France is"], max_tokens=8)
     assert output
     print(f"{output[0][1]}")

tests/quantization/test_experts_int8.py

@@ -34,6 +34,10 @@ def test_model_experts_int8_startup(
     model_info.check_transformers_version(on_fail="skip")
 
     with vllm_runner(
-        model, dtype=dtype, enforce_eager=True, quantization="experts_int8"
+        model,
+        dtype=dtype,
+        enforce_eager=True,
+        quantization="experts_int8",
+        allow_deprecated_quantization=True,
     ) as vllm_model:
         vllm_model.generate_greedy(example_prompts, max_tokens)

tests/quantization/test_rtn.py

@@ -30,6 +30,10 @@ def test_model_rtn_startup(
     max_tokens: int,
 ) -> None:
     with vllm_runner(
-        model, enforce_eager=True, dtype=dtype, quantization="rtn"
+        model,
+        enforce_eager=True,
+        dtype=dtype,
+        quantization="rtn",
+        allow_deprecated_quantization=True,
     ) as vllm_model:
         vllm_model.generate_greedy(example_prompts, max_tokens)