revert step hints for speech workloads

Author: priyakasimbeg

algoperf/workloads/librispeech_conformer/workload.py

@@ -89,6 +89,4 @@ def eval_period_time_sec(self) -> int:
   @property
   def step_hint(self) -> int:
     """Approx. steps the baseline can do in the allowed runtime budget."""
-    # TODO(kasimbeg):rever tot 76000
-    # return 76_000
-    return 80_000
+    return 76_000

algoperf/workloads/librispeech_deepspeech/librispeech_jax/workload.py

@@ -96,9 +96,7 @@ def test_target_value(self) -> float:
   @property
   def step_hint(self) -> int:
     """Approx. steps the baseline can do in the allowed runtime budget."""
-    # TODO(kasimbeg): revert old version
-    # return 38_400
-    return 48_000
+    return 38_400
 
   @property
   def max_allowed_runtime_sec(self) -> int: