remove useless logs (#55729)

python/paddle/distributed/fleet/meta_optimizers/dygraph_optimizer/dygraph_sharding_optimizer.py

@@ -23,10 +23,8 @@ from paddle.distributed import fleet
 from ...utils.log_util import logger
 from ...utils.tensor_fusion_helper import fused_parameters
 
-g_shard_use_reduce = int(os.environ.get("FLAGS_shard_use_reduce", 0))
-logger.info(f"g_shard_use_reduce {g_shard_use_reduce}")
-g_shard_norm_align_dp = int(os.environ.get("FLAGS_shard_norm_align_dp", 1))
-logger.info(f"g_shard_norm_align_dp {g_shard_norm_align_dp}")
+g_shard_use_reduce = int(os.environ.get("FLAGS_shard_use_reduce", 1))
+g_shard_norm_align_dp = int(os.environ.get("FLAGS_shard_norm_align_dp", 0))
 
 if g_shard_norm_align_dp:
     assert (

python/paddle/distributed/fleet/meta_optimizers/dygraph_optimizer/hybrid_parallel_optimizer.py

@@ -39,8 +39,7 @@ from ...utils.mix_precision_utils import MixPrecisionOptimizer
 
 __all__ = []
 
-g_shard_norm_align_dp = int(os.environ.get("FLAGS_shard_norm_align_dp", 1))
-logger.info(f"g_shard_norm_align_dp {g_shard_norm_align_dp}")
+g_shard_norm_align_dp = int(os.environ.get("FLAGS_shard_norm_align_dp", 0))
 
 
 class HybridParallelClipGrad:

python/paddle/distributed/fleet/meta_parallel/pipeline_parallel.py

@@ -44,8 +44,7 @@ from paddle.distributed.fleet.utils.tensor_fusion_helper import (
 
 __all__ = []
 
-g_shard_use_reduce = int(os.environ.get("FLAGS_shard_use_reduce", 0))
-logger.info(f"g_shard_use_reduce {g_shard_use_reduce}")
+g_shard_use_reduce = int(os.environ.get("FLAGS_shard_use_reduce", 1))
 
 
 # assume only the first stage and last stage need data, and data consumption is ordred