[Bugfix] Use AutoTune cache_input_tensors properly (#483)

tilelang/autotuner/__init__.py

@@ -249,6 +249,19 @@ class AutoTuner:
         if self.jit_compile is None:
             self.jit_compile = _compile
 
+        # Factory functions for generating input tensors.
+        # This encapsulates the logic of using either a custom supply program (`supply_prog`)
+        # or the default profiler input generation (`profiler._get_inputs`).
+        def get_input_tensors_supply(supply_prog, profiler, with_output: bool):
+
+            def func():
+                if supply_prog is not None:
+                    return supply_prog(profiler._get_params(with_output=with_output))
+                else:
+                    return profiler._get_inputs(with_output=with_output)
+
+            return func
+
         def target_fn(jit_context: JITContext):
             # Unpack the context
             kernel = jit_context.kernel
@@ -264,57 +277,30 @@ class AutoTuner:
 
             profiler = kernel.get_profiler(tensor_supply_type=supply_type)
 
-            # Factory functions for generating input tensors.
-            # This encapsulates the logic of using either a custom supply program (`supply_prog`)
-            # or the default profiler input generation (`profiler._get_inputs`).
-            def get_input_tensors_supply(with_output: bool):
-
-                def func():
-                    if supply_prog is not None:
-                        return supply_prog(profiler._get_params(with_output=with_output))
-                    else:
-                        return profiler._get_inputs(with_output=with_output)
-
-                return func
-
-            jit_input_tensors_supply = get_input_tensors_supply(with_output=False)
-            ref_input_tensors_supply = get_input_tensors_supply(with_output=False)
-
-            if cache_input_tensors:
-                jit_input_tensors = jit_input_tensors_supply()
-                if self.jit_input_tensors is not None:
-                    if not check_tensor_list_compatibility(self.jit_input_tensors,
-                                                           jit_input_tensors):
-                        logger.warning(
-                            "Incompatible input tensor properties detected between cached tensors and "
-                            "tensors regenerated for the current configuration trial. "
-                            "This can happen if different tuning configurations require different input shapes/dtypes "
-                            "and input tensor caching is enabled.\n"
-                            "To ensure fresh, compatible inputs are generated for every trial "
-                            "you can disable caching by setting:\n"
-                            "  `cache_input_tensors=False`\n"
-                            "within your `.set_compile_args(...)` call.\n")
-                    self.jit_input_tensors = jit_input_tensors
-                self.jit_input_tensors = jit_input_tensors
+            if cache_input_tensors and self.jit_input_tensors is not None:
+                jit_input_tensors = self.jit_input_tensors
             else:
-                self.jit_input_tensors = jit_input_tensors_supply()
+                jit_input_tensors_supply = get_input_tensors_supply(
+                    supply_prog, profiler, with_output=False)
+                jit_input_tensors = jit_input_tensors_supply()
 
             if (not skip_check) and (ref_prog is not None):
                 if manual_check_prog is not None:
                     profiler.manual_assert_close(
                         ref_prog,
-                        input_tensors=self.jit_input_tensors,
+                        input_tensors=jit_input_tensors,
                         manual_check_prog=manual_check_prog)
                 else:
                     profiler.assert_allclose(
                         ref_prog,
-                        input_tensors=self.jit_input_tensors,
+                        input_tensors=jit_input_tensors,
                         rtol=rtol,
                         atol=atol,
                         max_mismatched_ratio=max_mismatched_ratio)
-            latency = profiler.do_bench(
-                warmup=warmup, rep=rep, input_tensors=self.jit_input_tensors)
+            latency = profiler.do_bench(warmup=warmup, rep=rep, input_tensors=jit_input_tensors)
             if self.ref_latency_cache is None and ref_prog is not None:
+                ref_input_tensors_supply = get_input_tensors_supply(
+                    supply_prog, profiler, with_output=False)
                 self.ref_input_tensors = ref_input_tensors_supply()
                 self.ref_latency_cache = profiler.do_bench(
                     ref_prog, n_warmup=warmup, n_repeat=rep, input_tensors=self.ref_input_tensors)
@@ -367,6 +353,14 @@ class AutoTuner:
                 continue
 
         ref_latency = None
+        if results_with_configs[0][0].cache_input_tensors:
+            supply_prog = results_with_configs[0][0].supply_prog
+            supply_type = results_with_configs[0][0].supply_type
+            profiler = results_with_configs[0][0].kernel.get_profiler(
+                tensor_supply_type=supply_type)
+            jit_input_tensors_supply = get_input_tensors_supply(
+                supply_prog, profiler, with_output=False)
+            self.jit_input_tensors = jit_input_tensors_supply()
         progress_bar = tqdm(range(len(results_with_configs)), desc="Bench configurations")
         for i in progress_bar:
             jit_context, config = results_with_configs[i]