[CI/Build] drop support for Python 3.8 EOL (#8464)

Signed-off-by: Aaron Pham <contact@aarnphm.xyz>

tests/compile/piecewise/test_toy_llama.py

@@ -429,8 +429,8 @@ def benchmark():
     # print in tabular format
     print("batch size\teager mode\tfull cudagraph\tpiecewise cudagraph")
     for b in cudagraph_sizes:
-        print((f"{b}\t{eager_time[b]:.3f}\t{full_cudagraph_time[b]:.3f}"
-               f"\t{piecewise_cudagraph_time[b]:.3f}"))
+        print(f"{b}\t{eager_time[b]:.3f}\t{full_cudagraph_time[b]:.3f}"
+              f"\t{piecewise_cudagraph_time[b]:.3f}")
 
 
 if __name__ == "__main__":

tests/conftest.py

@@ -1,6 +1,5 @@
 import json
 import os
-import sys
 import tempfile
 from collections import UserList
 from enum import Enum
@@ -52,7 +51,7 @@ PromptVideoInput = _PromptMultiModalInput[np.ndarray]
 
 
 def _read_prompts(filename: str) -> List[str]:
-    with open(filename, "r") as f:
+    with open(filename) as f:
         prompts = f.readlines()
         return prompts
 
@@ -62,14 +61,8 @@ class _ImageAssetPrompts(TypedDict):
     cherry_blossom: str
 
 
-if sys.version_info < (3, 9):
-    # UserList cannot be subscripted
-    class _ImageAssetsBase(UserList):
-        pass
-else:
-
-    class _ImageAssetsBase(UserList[ImageAsset]):
-        pass
+class _ImageAssetsBase(UserList[ImageAsset]):
+    pass
 
 
 class _ImageAssets(_ImageAssetsBase):
@@ -94,14 +87,8 @@ class _VideoAssetPrompts(TypedDict):
     sample_demo_1: str
 
 
-if sys.version_info < (3, 9):
-    # UserList cannot be subscripted
-    class _VideoAssetsBase(UserList):
-        pass
-else:
-
-    class _VideoAssetsBase(UserList[VideoAsset]):
-        pass
+class _VideoAssetsBase(UserList[VideoAsset]):
+    pass
 
 
 class _VideoAssets(_VideoAssetsBase):
@@ -958,7 +945,7 @@ def dummy_opt_path():
                               "*.msgpack"
                           ])
         assert os.path.exists(json_path)
-        with open(json_path, "r") as f:
+        with open(json_path) as f:
             config = json.load(f)
         config["architectures"] = ["MyOPTForCausalLM"]
         with open(json_path, "w") as f:
@@ -977,7 +964,7 @@ def dummy_llava_path():
                               "*.msgpack"
                           ])
         assert os.path.exists(json_path)
-        with open(json_path, "r") as f:
+        with open(json_path) as f:
             config = json.load(f)
         config["architectures"] = ["MyLlava"]
         with open(json_path, "w") as f:
@@ -996,7 +983,7 @@ def dummy_gemma2_embedding_path():
                               "*.msgpack"
                           ])
         assert os.path.exists(json_path)
-        with open(json_path, "r") as f:
+        with open(json_path) as f:
             config = json.load(f)
         config["architectures"] = ["MyGemma2Embedding"]
         with open(json_path, "w") as f:

tests/core/block/test_prefix_caching_block.py

@@ -99,13 +99,11 @@ class TestPrefixCachingBlock:
 
         token_ids = [random.randint(0, 50_000) for _ in range(num_tokens)]
 
-        first_chain, second_chain = [
-            TestPrefixCachingBlock.create_chain(
-                block_size=block_size,
-                token_ids=token_ids,
-                num_empty_trailing_blocks=num_empty_trailing_blocks)
-            for _ in range(2)
-        ]
+        first_chain, second_chain = (TestPrefixCachingBlock.create_chain(
+            block_size=block_size,
+            token_ids=token_ids,
+            num_empty_trailing_blocks=num_empty_trailing_blocks)
+                                     for _ in range(2))
 
         for first_chain_block, second_chain_block in zip(
                 first_chain, second_chain):

tests/kernels/test_mamba_ssm.py

@@ -510,7 +510,7 @@ def test_selective_scan_varlen(with_padding, is_variable_B, is_variable_C,
         for var in (u_ref, delta_ref, B_ref, C_ref, z_ref)
     ]
     for i in range(len(seqlens[0])):
-        u_s, delta_s, B_s, C_s, z_s = [v[i].unsqueeze(0) for v in splits]
+        u_s, delta_s, B_s, C_s, z_s = (v[i].unsqueeze(0) for v in splits)
         if padded_state_indices[i] == PAD_SLOT_ID:
             continue
         out_ref_s, _ = selective_scan_ref(

tests/models/decoder_only/vision_language/mm_processor_kwargs/test_qwen.py

@@ -104,7 +104,7 @@ def test_input_mapper_valid_mm_data(input_mapper_for_qwen,
 # Sad path tests for the multimodal input processor and mapper, respectively
 @pytest.mark.parametrize("mm_data", [
     {
-        "image": torch.rand((5))
+        "image": torch.rand(5)
     },
     {
         "image": torch.rand((5, 5, 5, 5, 5))

tests/samplers/test_rejection_sampler.py

@@ -413,12 +413,10 @@ class _CorrectnessTestHelper:
     def generate_probs_for_test(
         self, draft_and_target_probs_equal: bool
     ) -> Tuple[torch.Tensor, torch.Tensor, torch.Tensor]:
-        draft_probs, target_probs = [
-            F.softmax(
-                torch.rand(self.vocab_size, dtype=torch.float32),
-                dim=-1,
-            ) for _ in range(2)
-        ]
+        draft_probs, target_probs = (F.softmax(
+            torch.rand(self.vocab_size, dtype=torch.float32),
+            dim=-1,
+        ) for _ in range(2))
 
         num_reference_probs = 100
         reference_probs = F.softmax(

tests/test_logger.py

@@ -29,7 +29,7 @@ def test_trace_function_call():
     cur_dir = os.path.dirname(__file__)
     enable_trace_function_call(path, cur_dir)
     f1(1)
-    with open(path, 'r') as f:
+    with open(path) as f:
         content = f.read()
 
     assert "f1" in content

tests/tokenization/test_detokenize.py

@@ -93,10 +93,10 @@ def test_mistral_edge_case(tokenizer, truth):
 def skip_special_tokens(request, tokenizer_name) -> Generator[bool, Any, None]:
     if "mistral" in tokenizer_name:
         yield (
-            bool(True) if request.param else
+            True if request.param else
             pytest.skip("mistral doesn't support skip_special_tokens=False"))
     else:
-        yield bool(True) if request.param else bool(False)
+        yield bool(request.param)
 
 
 @pytest.mark.parametrize("truth", TRUTH)