Re-enable transformers tests (mlflow#10343)

Signed-off-by: Serena Ruan <[email protected]>

tests/transformers/test_transformers_model_export.py

@@ -253,7 +253,7 @@ def fill_mask_pipeline():
 @flaky()
 def text2text_generation_pipeline():
     task = "text2text-generation"
-    architecture = "mrm8488/t5-base-finetuned-common_gen"
+    architecture = "mrm8488/t5-small-finetuned-common_gen"
     model = transformers.T5ForConditionalGeneration.from_pretrained(architecture)
     tokenizer = transformers.T5TokenizerFast.from_pretrained(architecture)
 
@@ -293,7 +293,7 @@ def translation_pipeline():
 @flaky()
 def summarizer_pipeline():
     task = "summarization"
-    architecture = "philschmid/distilbart-cnn-12-6-samsum"
+    architecture = "sshleifer/distilbart-cnn-6-6"
     model = transformers.BartForConditionalGeneration.from_pretrained(architecture)
     tokenizer = transformers.AutoTokenizer.from_pretrained(architecture)
     return transformers.pipeline(
@@ -363,7 +363,7 @@ def ner_pipeline_aggregation():
 @pytest.fixture
 @flaky()
 def conversational_pipeline():
-    return transformers.pipeline(model="microsoft/DialoGPT-medium")
+    return transformers.pipeline(model="AVeryRealHuman/DialoGPT-small-TonyStark")
 
 
 @pytest.fixture
@@ -1341,18 +1341,17 @@ def test_qa_pipeline_pyfunc_load_and_infer(small_qa_pipeline, model_path, infere
     assert all(isinstance(element, str) for element in inference)
 
 
-@pytest.mark.skipif(RUNNING_IN_GITHUB_ACTIONS, reason=GITHUB_ACTIONS_SKIP_REASON)
 @pytest.mark.parametrize(
     ("data", "result"),
     [
-        ("muppet keyboard type", ["A muppet is typing on a keyboard."]),
+        ("muppet keyboard type", ["A man is typing a muppet on a keyboard."]),
         (
             ["pencil draw paper", "pie apple eat"],
             # NB: The result of this test case, without inference config overrides is:
             # ["A man drawing on paper with pencil", "A man eating a pie with applies"]
             # The inference config override forces additional insertion of more grammatically
             # correct responses to validate that the inference config is being applied.
-            ["A man is drawing on paper with a pencil.", "A man is eating a pie with apples."],
+            ["A man draws a pencil on a paper.", "A man eats a pie of apples."],
         ),
     ],
 )
@@ -1510,7 +1509,6 @@ def test_text2text_generation_pipeline_with_params_with_errors(
         pyfunc_loaded.predict(data, {"top_k": "2"})
 
 
-@pytest.mark.skipif(RUNNING_IN_GITHUB_ACTIONS, reason=GITHUB_ACTIONS_SKIP_REASON)
 def test_text2text_generation_pipeline_with_inferred_schema(text2text_generation_pipeline):
     with mlflow.start_run():
         model_info = mlflow.transformers.log_model(
@@ -1519,7 +1517,7 @@ def test_text2text_generation_pipeline_with_inferred_schema(text2text_generation
     pyfunc_loaded = mlflow.pyfunc.load_model(model_info.model_uri)
 
     assert pyfunc_loaded.predict("muppet board nails hammer") == [
-        "A muppet is hammering nails on a board."
+        "A hammer with a muppet and nails on a board."
     ]
 
 
@@ -1530,7 +1528,6 @@ def test_text2text_generation_pipeline_with_inferred_schema(text2text_generation
         ([{"answer": ["42"], "context": "life"}, {"unmatched": "keys", "cause": "failure"}]),
     ],
 )
-@pytest.mark.skipif(RUNNING_IN_GITHUB_ACTIONS, reason=GITHUB_ACTIONS_SKIP_REASON)
 def test_invalid_input_to_text2text_pipeline(text2text_generation_pipeline, invalid_data):
     # Adding this validation test due to the fact that we're constructing the input to the
     # Pipeline. The Pipeline requires a format of a pseudo-dict-like string. An example of
@@ -1549,7 +1546,6 @@ def test_invalid_input_to_text2text_pipeline(text2text_generation_pipeline, inva
 @pytest.mark.parametrize(
     "data", ["Generative models are", (["Generative models are", "Computers are"])]
 )
-@pytest.mark.skipif(RUNNING_IN_GITHUB_ACTIONS, reason=GITHUB_ACTIONS_SKIP_REASON)
 def test_text_generation_pipeline(text_generation_pipeline, model_path, data):
     signature = infer_signature(
         data, mlflow.transformers.generate_signature_output(text_generation_pipeline, data)
@@ -1598,7 +1594,6 @@ def test_text_generation_pipeline(text_generation_pipeline, model_path, data):
         (["tell me a story", {"of": "a properly configured pipeline input"}]),
     ],
 )
-@pytest.mark.skipif(RUNNING_IN_GITHUB_ACTIONS, reason=GITHUB_ACTIONS_SKIP_REASON)
 def test_invalid_input_to_text_generation_pipeline(text_generation_pipeline, invalid_data):
     if isinstance(invalid_data, list):
         match = "If supplying a list, all values must be of string type"
@@ -1798,7 +1793,6 @@ def test_translation_pipeline(translation_pipeline, model_path, data, result):
         ["Baking cookies is quite easy", "Writing unittests is good for"],
     ],
 )
-@pytest.mark.skipif(RUNNING_IN_GITHUB_ACTIONS, reason=GITHUB_ACTIONS_SKIP_REASON)
 def test_summarization_pipeline(summarizer_pipeline, model_path, data):
     model_config = {
         "top_k": 2,
@@ -1923,7 +1917,6 @@ def test_ner_pipeline(pipeline_name, model_path, data, result, request):
     assert pd_inference == result
 
 
-@pytest.mark.skipif(RUNNING_IN_GITHUB_ACTIONS, reason=GITHUB_ACTIONS_SKIP_REASON)
 def test_conversational_pipeline(conversational_pipeline, model_path):
     signature = infer_signature(
         "Hi there!",
@@ -1935,21 +1928,21 @@ def test_conversational_pipeline(conversational_pipeline, model_path):
 
     first_response = loaded_pyfunc.predict("What is the best way to get to Antarctica?")
 
-    assert first_response == "I think you can get there by boat."
+    assert first_response == "The best way would be to go to space."
 
     second_response = loaded_pyfunc.predict("What kind of boat should I use?")
 
-    assert second_response == "A boat that can go to Antarctica."
+    assert second_response == "The best way to get to space would be to reach out and touch it."
 
     # Test that a new loaded instance has no context.
     loaded_again_pyfunc = mlflow.pyfunc.load_model(model_path)
     third_response = loaded_again_pyfunc.predict("What kind of boat should I use?")
 
-    assert third_response == "A boat that can't sink."
+    assert third_response == "The one with the guns."
 
     fourth_response = loaded_again_pyfunc.predict("Can I use it to go to the moon?")
 
-    assert fourth_response == "Only if you have a boat that can't sink."
+    assert fourth_response == "Sure."
 
 
 @pytest.mark.parametrize(
@@ -2902,7 +2895,6 @@ def test_instructional_pipeline_with_prompt_in_output(model_path):
         ),
     ],
 )
-@pytest.mark.skipif(RUNNING_IN_GITHUB_ACTIONS, reason=GITHUB_ACTIONS_SKIP_REASON)
 @pytest.mark.skipcacheclean
 def test_signature_inference(pipeline_name, data, result, request):
     pipeline = request.getfixturevalue(pipeline_name)