chore: update requires-python in pyproject.toml (#2080)

Author: raayandhar

benchmarks/bench_batch_attention.py

@@ -436,7 +436,7 @@ def main(args: argparse.Namespace) -> None:
     records_new = []
     records_separate = []
     for cfg_id, (decode_case, prefill_case) in enumerate(
-        zip(decode_lens, prefill_lens), start=1
+        zip(decode_lens, prefill_lens, strict=True), start=1
     ):
         prefill_kv_lens = [p[0] for p in prefill_case]
         prefill_qo_lens = [p[1] for p in prefill_case]

benchmarks/bench_mixed_attention.py

@@ -218,7 +218,7 @@ def _run_single_prefill():
     head_dim = 128
 
     for idx, (p_q_lens, p_kv_lens, d_q_len, d_kv_len) in enumerate(
-        zip(p_q_configs, p_kv_configs, d_q_len_configs, d_kv_len_configs)
+        zip(p_q_configs, p_kv_configs, d_q_len_configs, d_kv_len_configs, strict=True)
     ):
         print(f"===== Benchmark {idx + 1}: (kv_len, qo_len) set =====")
         run_bench(

docs/installation.rst

@@ -13,7 +13,7 @@ Prerequisites
 
 - OS: Linux only
 
-- Python: 3.9, 3.10, 3.11, 3.12, 3.13
+- Python: 3.10, 3.11, 3.12, 3.13, 3.14
 
 Quick Start
 ^^^^^^^^^^^

flashinfer/autotuner.py

@@ -648,7 +648,9 @@ def _generate_optimization_profiles(
 
             opt_shapes_max = {
                 v1: v2
-                for v1, v2 in zip(opt_shapes, tuple(opt_shapes[1:]) + (float("inf"),))
+                for v1, v2 in zip(
+                    opt_shapes, tuple(opt_shapes[1:]) + (float("inf"),), strict=True
+                )
             }
             dynamic_dims.append(
                 (spec.input_idx, spec.dim_idx, opt_shapes_max, opt_shapes)

flashinfer/cascade.py

@@ -349,6 +349,7 @@ def __init__(
                     paged_kv_indptr_buf_arr,
                     paged_kv_indices_buf_arr,
                     paged_kv_last_page_len_buf_arr,
+                    strict=True,
                 )
             ]
         else:
@@ -381,7 +382,7 @@ def reset_workspace_buffer(
             be the same as the device of the input tensors.
         """
         for wrapper, int_workspace_buffer in zip(
-            self._batch_prefill_wrappers, int_workspace_buffers
+            self._batch_prefill_wrappers, int_workspace_buffers, strict=True
         ):
             wrapper.reset_workspace_buffer(float_workspace_buffer, int_workspace_buffer)
 
@@ -479,6 +480,7 @@ def plan(
                 paged_kv_indptr_arr,
                 paged_kv_indices_arr,
                 paged_kv_last_page_len,
+                strict=True,
             )
         ):
             wrapper.plan(

flashinfer/comm/mnnvl.py

@@ -414,7 +414,7 @@ def open_mnnvl_memory(mapping: Mapping, size: int):
                 pidfds.append(pidfd)
 
             remote_fds = []
-            for pidfd, fd in zip(pidfds, all_handles_data):
+            for pidfd, fd in zip(pidfds, all_handles_data, strict=True):
                 remote_fd = syscall(SYS_pidfd_getfd, pidfd, fd, 0)
                 if remote_fd < 0:
                     err = ctypes.get_errno()

flashinfer/cute_dsl/blockscaled_gemm.py

@@ -154,6 +154,7 @@ def __new_from_mlir_values__(self, values):
                 self._cluster_shape_mnk,
             ],
             self._values_pos,
+            strict=True,
         ):
             obj_list.append(new_from_mlir_values(obj, values[:n_items]))
             values = values[n_items:]
@@ -348,6 +349,7 @@ def _get_current_work_for_linear_idx(
                 cur_cluster_coord,
                 self.cta_id_in_cluster,
                 (*self.params.cluster_shape_mn, Int32(1)),
+                strict=True,
             )
         )
 

flashinfer/jit/attention/utils.py

@@ -30,11 +30,14 @@ def generate_additional_params(
             for dtype, var in zip(
                 additional_tensor_dtypes,
                 additional_tensor_names,
+                strict=True,
             )
         ]
         + [
             f"{dtype} {var};\n"
-            for dtype, var in zip(additional_scalar_dtypes, additional_scalar_names)
+            for dtype, var in zip(
+                additional_scalar_dtypes, additional_scalar_names, strict=True
+            )
         ]
     )
     additional_func_params = "".join(
@@ -48,7 +51,9 @@ def generate_additional_params(
         ]
         + [
             f", {dtype} {var}"
-            for dtype, var in zip(additional_scalar_dtypes, additional_scalar_names)
+            for dtype, var in zip(
+                additional_scalar_dtypes, additional_scalar_names, strict=True
+            )
         ]
     )
     if is_sm90_template:
@@ -59,7 +64,9 @@ def generate_additional_params(
                     if var.startswith("maybe")
                     else f"params.additional_params.{var} = static_cast<{dtype}*>({var}.data_ptr());"
                 )
-                for dtype, var in zip(additional_tensor_dtypes, additional_tensor_names)
+                for dtype, var in zip(
+                    additional_tensor_dtypes, additional_tensor_names, strict=True
+                )
             ]
             + [
                 f"params.additional_params.{var} = {var};"
@@ -74,7 +81,9 @@ def generate_additional_params(
                     if var.startswith("maybe")
                     else f"params.{var} = static_cast<{dtype}*>({var}.data_ptr());"
                 )
-                for dtype, var in zip(additional_tensor_dtypes, additional_tensor_names)
+                for dtype, var in zip(
+                    additional_tensor_dtypes, additional_tensor_names, strict=True
+                )
             ]
             + [f"params.{var} = {var};" for var in additional_scalar_names]
         )

pyproject.toml

@@ -15,7 +15,7 @@
 [project]
 name = "flashinfer-python"
 description = "FlashInfer: Kernel Library for LLM Serving"
-requires-python = ">=3.9,<4.0"
+requires-python = ">=3.10,<4.0"
 authors = [{ name = "FlashInfer team" }]
 license = "Apache-2.0"
 readme = "README.md"