[DeepSpeed ZeRO3] Fix performance degradation in sharded models (#18911)

* [DeepSpeed] Fix performance degradation in sharded models

* style

* polish
Co-authored-by: Stas Bekman <stas@stason.org>

src/transformers/modeling_utils.py

@@ -421,10 +421,15 @@ def _load_state_dict_into_model(model_to_load, state_dict, start_prefix):
         if is_deepspeed_zero3_enabled():
             import deepspeed
 
+            # In sharded models, each shard has only part of the full state_dict, so only gather
+            # parameters that are in the current state_dict.
+            named_parameters = dict(module.named_parameters(prefix=prefix[:-1], recurse=False))
+            params_to_gather = [named_parameters[k] for k in state_dict.keys() if k in named_parameters]
+            if len(params_to_gather) > 0:
                 # because zero3 puts placeholders in model params, this context
                 # manager gathers (unpartitions) the params of the current layer, then loads from
                 # the state dict and then re-partitions them again
-            with deepspeed.zero.GatheredParameters(list(module.parameters(recurse=False)), modifier_rank=0):
+                with deepspeed.zero.GatheredParameters(params_to_gather, modifier_rank=0):
                     if torch.distributed.get_rank() == 0:
                         module._load_from_state_dict(*args)
         else: