add last_prologue_traces earlier (#917)

thunder/__init__.py

@@ -593,15 +593,22 @@ def get_computation_and_inputs(*args, **kwargs):
                 executors_list=(pythonex,),
                 use_del_last_used=False,
             )
-            protrace = prologue_traces[-1]
-            pro = protrace.python_callable()
+            prologue_trc = prologue_traces[-1]
+            pro = prologue_trc.python_callable()
 
             if epilogue_trc is not None:
                 epilogue = epilogue_trc.python_callable()
             else:
                 epilogue = None
 
             cs.last_prologue_transformation_stop = time.perf_counter_ns()
+            cs.last_prologue_traces = prologue_traces
+            cs.last_prologue = pro
+            cs.last_traces = computation_traces
+            backward_traces = []
+            cs.last_backward_traces = backward_traces
+            cs.last_interpreter_log = last_interpreter_log
+            cs.last_interpreted_instructions = (i for i in last_interpreter_log if isinstance(i, dis.Instruction))
 
             cs.last_prologue_execution_start = time.perf_counter_ns()
             if interpretation is INTERPRETATION_OPTIONS.TRANSLATE_PYTHON:
@@ -611,12 +618,6 @@ def get_computation_and_inputs(*args, **kwargs):
                 pro_to_epi = None
             cs.last_prologue_execution_stop = time.perf_counter_ns()
 
-            cs.last_traces = computation_traces
-            backward_traces = []
-            cs.last_backward_traces = backward_traces
-            cs.last_interpreter_log = last_interpreter_log
-            cs.last_interpreted_instructions = (i for i in last_interpreter_log if isinstance(i, dis.Instruction))
-
             computation_trc = dce(computation_trc)
             computation_traces.append(computation_trc)
 
@@ -633,7 +634,6 @@ def get_computation_and_inputs(*args, **kwargs):
                     computation_trc, backward_trc = split_forward_backward(computation_trc, cd, cs, *inps)
                     # Note computation_trc and backward_trc have been appended to cs.last_(backward_)traces
                     # by split_forward_backward
-                    extraces = cs.last_traces
 
             if backward_trc is None:
                 ## EPILOGUE and TRANSFORMS should not mix...
@@ -666,14 +666,14 @@ def get_computation_and_inputs(*args, **kwargs):
                         executors_list=cd.executors_list,
                         use_del_last_used=False,
                     )
-                computation_traces.append(computation_trc)
-                computation_trc = extraces[-1]
+                computation_traces.extend(extraces)
+                computation_trc = computation_traces[-1]
 
             if cd.use_cudagraphs:
                 from thunder.executors.cudagraphex import cudagraphex
 
                 computation_trc = cudagraphex.fusion_pass(computation_trc)
-                extraces.append(computation_trc)
+                computation_traces.append(computation_trc)
 
                 if backward_trc is not None:
                     backward_trc = cudagraphex.fusion_pass(backward_trc, num_static_inputs=len(backward_trc.args[0][0]))
@@ -684,7 +684,7 @@ def get_computation_and_inputs(*args, **kwargs):
 
             if not compile_options.get("disable_inplace_copy_check", False):
                 thunder.core.transform_common._inplace_copy_sanity_check(computation_trc)
-                extraces.append(computation_trc)
+                computation_traces.append(computation_trc)
 
             for transform in transforms:
                 # NOTE: `backward_trc` could be None.
@@ -693,7 +693,7 @@ def get_computation_and_inputs(*args, **kwargs):
                 )
                 if new_computation_trc is not computation_trc:
                     computation_trc = new_computation_trc
-                    extraces.append(computation_trc)
+                    computation_traces.append(computation_trc)
                 if backward_trc is not None:
                     new_backward_trc = transform.transform_trace_post_optimization(
                         backward_trc, executors_list=cd.executors_list
@@ -715,7 +715,7 @@ def get_computation_and_inputs(*args, **kwargs):
                 pro,
                 prologue_traces,
                 comp,
-                extraces,
+                computation_traces,
                 epilogue,
                 epilogue_traces,
                 backward_fn,
@@ -726,10 +726,6 @@ def get_computation_and_inputs(*args, **kwargs):
             if cd.cache_option is not CACHE_OPTIONS.NO_CACHING:
                 cs.interpreter_cache.append(cache_entry)
 
-            cs.last_traces += extraces
-            cs.last_prologue_traces = [prologue_trc] + prologue_traces
-            cs.last_prologue = pro
-
         return cache_entry, inps, pro_to_epi
 
     cd.get_computation_and_inputs = get_computation_and_inputs

thunder/tests/test_jit_general.py

@@ -1203,3 +1203,20 @@ def fn():
         return torch.cuda.get_device_properties(0).major
 
     assert fn() == thunder.jit(fn)()
+
+
+def test_failing_prologue_in_last_prologue_traces():
+    # we know that this will fail in the prologue
+    i = 0
+
+    def fn():
+        nonlocal i
+        i += 1
+        return i
+
+    jfn = thunder.jit(fn)
+    with pytest.raises(RuntimeError, match="Expected 1 to be equal to and have the type of 0"):
+        jfn()
+
+    # make sure that we have prologue traces in the last_prologue_traces
+    assert len(thunder.last_prologue_traces(jfn)) > 0