Fix memory leak caused by context parallelism hanging references by o…

…megaconf (NVIDIA#8299) (NVIDIA#8319) * save cp_size to self * use parallel_state instead of self --------- Signed-off-by: Jimmy Zhang <[email protected]> Co-authored-by: JimmyZhang12 <[email protected]> Co-authored-by: Jimmy Zhang <[email protected]> Co-authored-by: Eric Harper <[email protected]>
rohitrango · Feb 5, 2024 · 6b57b76 · 6b57b76
1 parent 97d1dc0
commit 6b57b76
Showing 1 changed file with 3 additions and 4 deletions.
diff --git a/nemo/collections/nlp/models/language_modeling/megatron_gpt_model.py b/nemo/collections/nlp/models/language_modeling/megatron_gpt_model.py
@@ -910,11 +910,11 @@ def get_batch(self, data_iterator, tuning):
         return batch
 
     def get_batch_on_this_context_parallel_rank(self, batch):
-        cp_size = self.cfg.get('context_parallel_size', 1)
         num_valid_tokens_in_ub = None
         if 'loss_mask' in batch and batch['loss_mask'] is not None:
             num_valid_tokens_in_ub = batch['loss_mask'].sum()
 
+        cp_size = parallel_state.get_context_parallel_world_size()
         if cp_size > 1:
             cp_rank = parallel_state.get_context_parallel_rank()
             for key, val in batch.items():
@@ -1011,7 +1011,7 @@ def fwd_output_and_loss_func(dataloader_iter, model, checkpoint_activations_all_
             def loss_func(output_tensor):
                 # Loss for a micro-batch (ub)
                 loss_for_ub = self.loss_func(batch['loss_mask'], batch['num_valid_tokens_in_ub'], output_tensor)
-                cp_size = self.cfg.get('context_parallel_size', 1)
+                cp_size = parallel_state.get_context_parallel_world_size()
                 if validation_step and not self.cfg.data.get('validation_drop_last', True):
                     num_valid_tokens_in_ub = batch['num_valid_tokens_in_ub']
                     if loss_for_ub.isnan():
@@ -1167,8 +1167,7 @@ def loss_func(self, loss_mask, num_valid_tokens_in_ub, output_tensor):
         loss_mask = loss_mask.view(-1).float()
         # TODO: add nemo version here
         loss = torch.sum(losses.view(-1) * loss_mask) / num_valid_tokens_in_ub  # sequence level nll
-        cp_size = self.cfg.get('context_parallel_size', 1)
-        if cp_size > 1:
+        if parallel_state.get_context_parallel_world_size() > 1:
             torch.distributed.all_reduce(loss, group=parallel_state.get_context_parallel_group())
         return loss