addressed comments

megatron/text_generation_utils.py

@@ -466,6 +466,8 @@ def sample_sequence_batch(model, context_tokens, context_lengths,
     with torch.no_grad():
         context_length = context_lengths.min().item()
 
+        # added eos_id to support the function generate_samples_eval that passes
+        # eos_id as an argument and needs termination when that id id found.
         if hasattr(args, 'eos_id'):
             eos_id = args.eos_id
         else: