[air/benchmarks] Fix typo in tensorflow_benchmark.py script preventin…

…g proper error surfacing (ray-project#32269)

There is a small typo in the tensorflow_benchmark.py script that does not properly catch when a vanilla TF run failed three times. Because of this, we would previously record a training time of 0.0 for vanilla TF, which skews the calculated average and suggests that vanilla TF outperformed Ray Train. Instead, we should have raised an error message to surface the problem.

Signed-off-by: Kai Fricke <kai@anyscale.com>
Signed-off-by: Edward Oakes <ed.nmi.oakes@gmail.com>

release/air_tests/air_benchmarks/workloads/tensorflow_benchmark.py

@@ -310,7 +310,7 @@ def run(
  config=config,
  )
  except Exception as e:
- if i > +2:
+ if i >= 2:
  raise RuntimeError("Vanilla TF run failed 3 times") from e
  print("Vanilla TF run failed:", e)
  continue
@@ -338,6 +338,7 @@ def run(
  times_ray.append(time_ray)
  times_local_ray.append(time_local_ray)
  losses_ray.append(loss_ray)
+
  times_vanilla.append(time_vanilla)
  times_local_vanilla.append(time_local_vanilla)
  losses_vanilla.append(loss_vanilla)