fix num_microbatches input for PP

ghstack-source-id: 459e8bee48fd77b027fecdc9e5f78ec375b87cb6
Pull Request resolved: #781

torchtitan/parallelisms/pipeline_llama.py

@@ -64,9 +64,7 @@ def pipeline_llama_manual_split(
     """
     pp_rank = pp_mesh.get_local_rank()
     pp_size = pp_mesh.size()
-    microbatches = (
-        job_config.experimental.pipeline_parallel_microbatches or parallel_dims.pp
-    )
+
     splits = (
         job_config.experimental.pipeline_parallel_split_points
         or generate_split_points(job_config, parallel_dims.pp, model_config)
@@ -117,7 +115,7 @@ def _build_stage(stage_idx, start_layer, stop_layer, is_first=False, is_last=Fal
         )
         logger.info(
             f"PP rank {pp_rank} is building stage_idx {stage_idx}"
-            f" with start_layer {start_layer}, stop_layer {stop_layer}: model chunk \n{model_chunk}"
+            f" with start_layer {start_layer}, stop_layer {stop_layer}"
         )
         stages.append(stage)
         models.append(model_chunk)

torchtitan/parallelisms/pipelining_utils.py

@@ -4,18 +4,27 @@
 # This source code is licensed under the BSD-style license found in the
 # LICENSE file in the root directory of this source tree.
 import os
-from typing import Tuple
+from typing import List, Tuple
 
 from torch.distributed.pipelining.schedules import (
     _PipelineScheduleRuntime,
     get_schedule_class,
     PipelineScheduleMulti,
     PipelineScheduleSingle,
 )
+from torchtitan.config_manager import JobConfig
 from torchtitan.logging import logger
+from torchtitan.models.llama.model import ModelArgs
 
 
-def generate_split_points(job_config, pp_dim, model_config):
+def generate_split_points(
+    job_config: JobConfig, pp_dim: int, model_config: ModelArgs
+) -> List[str]:
+    """
+    Generate a default split point based on the number of layers and
+    pipeline parallel dimension.
+    """
+
     schedule_class = get_schedule_class(
         job_config.experimental.pipeline_parallel_schedule
     )
@@ -51,7 +60,7 @@ def generate_split_points(job_config, pp_dim, model_config):
                 current_layer += base_interval
         splits.append("layers." + str(current_layer))
     logger.info(
-        f"No 'pipeline_parallel_split_points' so the generated splits are: {splits} \
+        f"No 'pipeline_parallel_split_points' provided so the generated splits are: {splits} \
 This may be sub-optimal as the number of layers per stage may be unbalanced."
     )
     return splits
@@ -73,18 +82,26 @@ def build_pipeline_schedule(job_config, stages, loss_fn):
         )
 
     looped_schedule = issubclass(schedule_class, PipelineScheduleMulti)
-    logger.info(
-        f"Using pipeline schedule {job_config.experimental.pipeline_parallel_schedule}"
-    )
     n_microbatches = job_config.experimental.pipeline_parallel_microbatches
+    # We expect that the number of local stages (`len(stages)`) is the same across all ranks
+    num_total_stages = job_config.experimental.pipeline_parallel_degree * len(stages)
     if n_microbatches is None:
-        n_microbatches = job_config.experimental.pipeline_parallel_degree
+        n_microbatches = num_total_stages
+    elif n_microbatches < num_total_stages:
+        logger.warning(
+            f"Number of microbatches ({n_microbatches}) is less than the total number \
+of stages ({num_total_stages}) which may result in a bubble in the pipeline."
+        )
 
     schedule = schedule_class(
         stages if looped_schedule else stages[0],
         n_microbatches=n_microbatches,
         loss_fn=loss_fn,
     )
+    logger.info(
+        f"Using pipeline schedule {job_config.experimental.pipeline_parallel_schedule} \
+with {n_microbatches} and {num_total_stages} stages."
+    )
 
     if pp_schedule_csv:
         assert schedule_class in [