only produce tensorboard logs on rank 0 by default

ghstack-source-id: 1d228f271db275dd229fae61b3ca064141afcacb
Pull Request resolved: #339

.ci/docker/requirements.txt

@@ -1,4 +1,3 @@
-torch >= 2.2.0.dev
 datasets
 tomli >= 1.1.0 ; python_version < "3.11"
 tensorboard

torchtitan/config_manager.py

@@ -113,6 +113,12 @@ def __init__(self):
             default="tb",
             help="Folder to dump TensorBoard states",
         )
+        self.parser.add_argument(
+            "--metrics.rank_0_only",
+            default=True,
+            action="store_true",
+            help="Whether to save TensorBoard metrics only for rank 0 or for all ranks",
+        )
 
         # model configs
         self.parser.add_argument(

torchtitan/metrics.py

@@ -113,16 +113,21 @@ def close(self):
 
 def build_metric_logger(config: JobConfig, tag: Optional[str] = None):
     dump_dir = config.job.dump_folder
-    save_tb_folder = config.metrics.save_tb_folder
-    # since we don't have run id yet, use current minute as identifier
+    tb_config = config.metrics
+    save_tb_folder = tb_config.save_tb_folder
+    # since we don't have run id, use current minute as the identifier
     datetime_str = datetime.now().strftime("%Y%m%d-%H%M")
     log_dir = os.path.join(dump_dir, save_tb_folder, datetime_str)
 
-    enable_tb = config.metrics.enable_tensorboard
+    enable_tb = tb_config.enable_tensorboard
     if enable_tb:
         logger.info(
             f"Metrics logging active. Tensorboard logs will be saved at {log_dir}"
         )
+        if tb_config.rank_0_only:
+            enable_tb = torch.distributed.get_rank() == 0
+        else:
+            rank_str = f"rank_{torch.distributed.get_rank()}"
+            log_dir = os.path.join(log_dir, rank_str)
 
-    rank_str = f"rank_{torch.distributed.get_rank()}"
-    return MetricLogger(os.path.join(log_dir, rank_str), tag, enable_tb)
+    return MetricLogger(log_dir, tag, enable_tb)

torchtitan/parallelisms/parallelize_llama.py

@@ -174,6 +174,7 @@ def parallelize_llama(model, world_mesh, parallel_dims, job_config: JobConfig):
         # Apply tensor + sequence parallelism to every transformer block
         for layer_id, transformer_block in model.layers.items():
             layer_plan = {
+                "attention_norm": SequenceParallel(),
                 "attention": PrepareModuleInput(
                     input_layouts=(Shard(1), None),
                     desired_input_layouts=(Replicate(), None),
@@ -182,15 +183,14 @@ def parallelize_llama(model, world_mesh, parallel_dims, job_config: JobConfig):
                 "attention.wk": col_parallel_strategy(),
                 "attention.wv": col_parallel_strategy(),
                 "attention.wo": row_parallel_strategy(output_layouts=Shard(1)),
-                "attention_norm": SequenceParallel(),
+                "ffn_norm": SequenceParallel(),
                 "feed_forward": PrepareModuleInput(
                     input_layouts=(Shard(1),),
                     desired_input_layouts=(Replicate(),),
                 ),
                 "feed_forward.w1": col_parallel_strategy(),
                 "feed_forward.w2": row_parallel_strategy(output_layouts=Shard(1)),
                 "feed_forward.w3": col_parallel_strategy(),
-                "ffn_norm": SequenceParallel(),
             }
 
             # Adjust attention module to use the local number of heads

train_configs/debug_model.toml

@@ -3,7 +3,6 @@
 [job]
 dump_folder = "./outputs"
 description = "Llama 3 debug training"
-# TODO: turn this back on once ci have tokenizer
 use_for_integration_test = true
 
 [profiling]
@@ -48,7 +47,7 @@ interval_type = "steps"
 interval = 5
 model_weights_only = false
 export_dtype = "float32"
-async_mode = "disabled" # ["disabled", "async", "async_with_pinned_mem"]
+async_mode = "disabled"  # ["disabled", "async", "async_with_pinned_mem"]
 
 [activation_checkpoint]
 mode = 'selective'  # ['none', 'selective', 'full']