[Test] Make model tests run again and remove --forked from pytest (#3631)

Co-authored-by: Simon Mo <simon.mo@hey.com>

tests/models/test_models.py

@@ -1,32 +1,28 @@
 """Compare the outputs of HF and vLLM when using greedy sampling.
 
-Run `pytest tests/models/test_models.py --forked`.
+This test only tests small models. Big models such as 7B should be tested from
+test_big_models.py because it could use a larger instance to run tests.
+
+Run `pytest tests/models/test_models.py`.
 """
 import pytest
 
 MODELS = [
     "facebook/opt-125m",
-    "meta-llama/Llama-2-7b-hf",
-    "mistralai/Mistral-7B-v0.1",
-    "Deci/DeciLM-7b",
-    "tiiuae/falcon-7b",
     "gpt2",
     "bigcode/tiny_starcoder_py",
-    "EleutherAI/gpt-j-6b",
     "EleutherAI/pythia-70m",
     "bigscience/bloom-560m",
-    "mosaicml/mpt-7b",
     "microsoft/phi-2",
     "stabilityai/stablelm-3b-4e1t",
-    "allenai/OLMo-1B",
+    # "allenai/OLMo-1B",  # Broken
     "bigcode/starcoder2-3b",
-    "Qwen/Qwen1.5-0.5B",
 ]
 
 
 @pytest.mark.parametrize("model", MODELS)
-@pytest.mark.parametrize("dtype", ["half"])
-@pytest.mark.parametrize("max_tokens", [128])
+@pytest.mark.parametrize("dtype", ["float"])
+@pytest.mark.parametrize("max_tokens", [96])
 def test_models(
     hf_runner,
     vllm_runner,
@@ -35,6 +31,9 @@ def test_models(
     dtype: str,
     max_tokens: int,
 ) -> None:
+    # To pass the small model tests, we need full precision.
+    assert dtype == "float"
+
     hf_model = hf_runner(model, dtype=dtype)
     hf_outputs = hf_model.generate_greedy(example_prompts, max_tokens)
     del hf_model